{"id":2103,"date":"2026-08-03T20:03:24","date_gmt":"2026-08-03T20:03:24","guid":{"rendered":"https:\/\/convly.ai\/?p=2103"},"modified":"2026-08-03T20:03:24","modified_gmt":"2026-08-03T20:03:24","slug":"what-is-gguf-format-explained","status":"publish","type":"post","link":"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/","title":{"rendered":"\u0645\u0627 \u0647\u0648 \u062a\u0646\u0633\u064a\u0642 GGUF\u061f \u0634\u0631\u062d \u062a\u0646\u0633\u064a\u0642 \u0645\u0644\u0641\u0627\u062a \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0627\u0644\u0630\u064a \u064a\u0639\u062a\u0645\u062f \u0639\u0644\u064a\u0647 \u0643\u0644\u064c\u0651 \u0645\u0646 Ollama \u0648llama.cpp"},"content":{"rendered":"<div class=\"convly-tldr\">\n<ul>\n<li><strong>GGUF \u0647\u0648 \u062a\u0646\u0633\u064a\u0642 \u0645\u0644\u0641\u0627\u062a\u064d \u064a\u064f\u0633\u062a\u062e\u062f\u0645 \u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0645\u062d\u0644\u064a\u064b\u0651\u0627.<\/strong> It packs a model&#8217;s weights, tokenizer and configuration into a single binary file that llama.cpp, Ollama, LM Studio and most other local-LLM tools load directly.<\/li>\n<li><strong>\u0648\u0642\u062f \u062d\u0644\u064e\u0651 \u0647\u0630\u0627 \u0627\u0644\u062a\u0646\u0633\u064a\u0642 \u0645\u062d\u0644\u064e\u0651 GGML \u0641\u064a \u0623\u063a\u0633\u0637\u0633 \u0662\u0660\u0662\u0663<\/strong> \u0644\u0623\u0646 \u0645\u0644\u0641\u0627\u062a GGML \u0643\u0627\u0646\u062a \u062a\u062a\u0648\u0642\u0641 \u0639\u0646 \u0627\u0644\u0639\u0645\u0644 \u0641\u0648\u0631 \u062a\u063a\u064a\u064f\u0651\u0631 \u0627\u0644\u062a\u0646\u0633\u064a\u0642. \u0623\u0645\u0627 GGUF \u0641\u0647\u0648 \u062a\u0646\u0633\u064a\u0642 \u0645\u064f\u0631\u0642\u064e\u0651\u0645 \u0648\u0642\u0627\u0628\u0644 \u0644\u0644\u062a\u0648\u0633\u064f\u0651\u0639\u060c \u0648\u0628\u0627\u0644\u062a\u0627\u0644\u064a \u062a\u0638\u0644 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0627\u0644\u0642\u062f\u064a\u0645\u0629 \u062a\u0639\u0645\u0644 \u062f\u0648\u0646 \u0645\u0634\u0627\u0643\u0644.<\/li>\n<li><strong>\u0648\u0627\u0644\u0644\u0627\u062d\u0642\u0629 \u062a\u0634\u064a\u0631 \u0625\u0644\u0649 \u062f\u0631\u062c\u0629 \u0627\u0644\u062a\u0643\u0645\u064a\u0645 (quantisation).<\/strong> Q4_K_M (\u062d\u0648\u0627\u0644\u064a \u0664,\u0669 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a \u0644\u0646\u0645\u0648\u0630\u062c \u0628\u062d\u062c\u0645 \u0668 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629) \u0647\u0648 \u0627\u0644\u0625\u0639\u062f\u0627\u062f \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a \u0627\u0644\u0642\u064a\u0627\u0633\u064a \u0645\u0646 \u062d\u064a\u062b \u0627\u0644\u062c\u0648\u062f\u0629 \u0648\u0627\u0644\u062d\u062c\u0645\u061b \u0628\u064a\u0646\u0645\u0627 Q8_0 \u064a\u0642\u062a\u0631\u0628 \u0645\u0646 \u0639\u062f\u0645 \u0641\u0642\u062f\u0627\u0646 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a \u062a\u0645\u0627\u0645\u064b\u0627 \u0639\u0646\u062f \u062d\u062c\u0645 \u062d\u0648\u0627\u0644\u064a \u0668,\u0665 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a\u061b \u0623\u0645\u0627 F16 \u0641\u0647\u0648 \u062a\u0646\u0633\u064a\u0642 \u063a\u064a\u0631 \u0645\u0643\u0645\u064e\u0651\u0645.<\/li>\n<li><strong>\u0642\u0627\u0639\u062f\u0629 \u0639\u0627\u0645\u0629:<\/strong> \u0627\u062e\u062a\u0631 \u0623\u0643\u0628\u0631 \u062f\u0631\u062c\u0629 \u062a\u0643\u0645\u064a\u0645 \u064a\u0633\u0645\u062d \u062d\u062c\u0645 \u0645\u0644\u0641\u0647\u0627 \u0628\u0627\u0644\u062a\u0646\u0627\u0633\u0628 \u0636\u0645\u0646 \u0633\u0639\u0629 \u0630\u0627\u0643\u0631\u0629 VRAM \u0644\u062f\u064a\u0643 \u0645\u0639 \u062a\u0631\u0643 \u0647\u0627\u0645\u0634 \u0661\u2013\u0662 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a \u0625\u0636\u0627\u0641\u064a \u0644\u0644\u0633\u064a\u0627\u0642.<\/li>\n<\/ul>\n<\/div>\n<p>GGUF (\u0648\u064a\u064f\u0634\u0627\u0631 \u0625\u0644\u064a\u0647 \u0639\u0627\u062f\u0629\u064b \u0628\u0627\u0633\u0645 GPT-Generated Unified Format) \u0647\u0648 \u062a\u0646\u0633\u064a\u0642 \u0645\u0644\u0641 \u062b\u0646\u0627\u0626\u064a \u0644\u062a\u062e\u0632\u064a\u0646 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629: \u0623\u064a \u0623\u0648\u0632\u0627\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0648\u0645\u064f\u0641\u0643\u0650\u0651\u0643 \u0627\u0644\u0631\u0645\u0648\u0632 \u0648\u062c\u0645\u064a\u0639 \u0628\u064a\u0627\u0646\u0627\u062a \u0627\u0644\u0625\u0639\u062f\u0627\u062f\u0627\u062a \u0627\u0644\u062a\u0643\u0648\u064a\u0646\u064a\u0629 \u0641\u064a \u0645\u0644\u0641 \u0648\u0627\u062d\u062f \u0645\u062a\u0643\u0627\u0645\u0644 \u0630\u0627\u062a\u064a\u064b\u0627. \u0648\u0647\u0648 \u0627\u0644\u062a\u0646\u0633\u064a\u0642 \u0627\u0644\u0623\u0635\u0644\u064a \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645 \u0641\u064a <strong>llama.cpp<\/strong>, \u0648\u0647\u0648 \u0645\u0627 \u062a\u0633\u062a\u062e\u062f\u0645\u0647 \u0628\u0631\u0627\u0645\u062c \u0645\u062b\u0644 <strong>Ollama<\/strong>, <strong>LM Studio<\/strong>KoboldCpp \u0648Jan \u0648\u0645\u0639\u0638\u0645 \u062a\u0637\u0628\u064a\u0642\u0627\u062a \u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0645\u062d\u0644\u064a\u064b\u0651\u0627 \u0627\u0644\u0623\u062e\u0631\u0649 \u0641\u0639\u0644\u064a\u064b\u0651\u0627. \u0641\u0625\u0630\u0627 \u0643\u0646\u062a \u0642\u062f \u0642\u0645\u062a \u064a\u0648\u0645\u064b\u0627 \u0628\u062a\u0646\u0632\u064a\u0644 \u0645\u0644\u0641 \u064a\u0646\u062a\u0647\u064a \u0627\u0645\u062a\u062f\u0627\u062f\u0647 \u0628\u0640 <code>.gguf<\/code>\u0623\u0648 \u0627\u0633\u062a\u0631\u062c\u0639\u062a \u0646\u0645\u0648\u0630\u062c\u064b\u0627 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0623\u0645\u0631 <code>ollama run<\/code>\u060c \u0641\u0623\u0646\u062a \u062d\u064a\u0646\u0647\u0627 \u062a\u0633\u062a\u062e\u062f\u0645 \u0647\u0630\u0627 \u0627\u0644\u062a\u0646\u0633\u064a\u0642.<\/p>\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_85 counter-flat ez-toc-counter ez-toc-container-direction\">\n<label for=\"ez-toc-cssicon-toggle-item-6a7128070c4e8\" class=\"ez-toc-cssicon-toggle-label\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">\u062a\u0628\u062f\u064a\u0644<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #000000;color:#000000\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #000000;color:#000000\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewbox=\"0 0 24 24\" version=\"1.2\" baseprofile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/label><input type=\"checkbox\"  id=\"ez-toc-cssicon-toggle-item-6a7128070c4e8\"  aria-label=\"\u062a\u0628\u062f\u064a\u0644\" \/><nav><ul class='ez-toc-list ez-toc-list-level-1' ><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#What_a_GGUF_file_actually_contains\" >\u0627\u0644\u0645\u062d\u062a\u0648\u0649 \u0627\u0644\u0641\u0639\u0644\u064a \u0644\u0645\u0644\u0641 GGUF<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#Why_GGUF_replaced_GGML\" >\u0627\u0644\u0633\u0628\u0628 \u0627\u0644\u0630\u064a \u062f\u0641\u0639 \u0625\u0644\u0649 \u0627\u0633\u062a\u0628\u062f\u0627\u0644 GGML \u0628\u0640 GGUF<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#How_to_read_the_quantisation_suffixes\" >\u0643\u064a\u0641\u064a\u0629 \u0642\u0631\u0627\u0621\u0629 \u0644\u0648\u0627\u062d\u0642 \u0627\u0644\u062a\u0643\u0645\u064a\u0645<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#Picking_a_quantisation_for_your_VRAM\" >\u0627\u062e\u062a\u064a\u0627\u0631 \u062f\u0631\u062c\u0629 \u062a\u0643\u0645\u064a\u0645 \u0645\u0646\u0627\u0633\u0628\u0629 \u0644\u0633\u0639\u0629 \u0630\u0627\u0643\u0631\u0629 VRAM \u0644\u062f\u064a\u0643<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#GGUF_vs_safetensors\" >\u0645\u0642\u0627\u0631\u0646\u0629 \u0628\u064a\u0646 GGUF \u0648safetensors<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#Where_GGUF_files_come_from\" >\u0645\u0635\u062f\u0631 \u0645\u0644\u0641\u0627\u062a GGUF<\/a><\/li><li class='ez-toc-page-1'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/convly.ai\/ar\/what-is-gguf-format-explained\/#Frequently_asked_questions\" >\u0627\u0644\u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u0634\u0627\u0626\u0639\u0629<\/a><\/li><\/ul><\/nav><\/div>\n<h2><span class=\"ez-toc-section\" id=\"What_a_GGUF_file_actually_contains\"><\/span>\u0627\u0644\u0645\u062d\u062a\u0648\u0649 \u0627\u0644\u0641\u0639\u0644\u064a \u0644\u0645\u0644\u0641 GGUF<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u064a\u0628\u062f\u0623 \u0645\u0644\u0641 GGUF \u0628\u0623\u0631\u0628\u0639\u0629 \u0628\u0627\u064a\u062a\u0627\u062a ASCII \u0647\u064a <code>GGUF<\/code>\u060c \u064a\u0644\u064a\u0647\u0627 \u0631\u0642\u0645 \u0627\u0644\u0625\u0635\u062f\u0627\u0631\u060c \u062b\u0645 \u0642\u0633\u0645 \u0644\u0644\u0645etadata \u0627\u0644\u0645\u0643\u0648\u064e\u0651\u0646 \u0645\u0646 \u0623\u0632\u0648\u0627\u062c \u0627\u0644\u0645\u0641\u062a\u0627\u062d-\u0627\u0644\u0642\u064a\u0645\u0629\u060c \u062b\u0645 \u062a\u062a\u0628\u0639\u0647 \u0627\u0644\u0645\u062a\u062c\u0647\u0627\u062a (tensors) \u0646\u0641\u0633\u0647\u0627. \u0648\u064a\u0639\u062a\u0628\u0631 \u0642\u0633\u0645 metadata \u0647\u0648 \u0627\u0644\u0642\u0631\u0627\u0631 \u0627\u0644\u062a\u0635\u0645\u064a\u0645\u064a \u0627\u0644\u0645\u0647\u0645: \u0641\u0647\u0648 \u064a\u062e\u0632\u0646 \u0628\u0646\u064a\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0648\u0637\u0648\u0644 \u0627\u0644\u0633\u064a\u0627\u0642 \u0648\u0627\u0644\u0645\u0641\u0631\u062f\u0627\u062a \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0645\u064f\u0641\u0643\u0650\u0651\u0643 \u0627\u0644\u0631\u0645\u0648\u0632 (tokenizer) \u0648\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0645\u062d\u0627\u062f\u062b\u0629 (chat template) \u0648\u062a\u0641\u0627\u0635\u064a\u0644 \u0627\u0644\u062a\u0643\u0645\u064a\u0645 \u0643\u0645\u0641\u0627\u062a\u064a\u062d \u0645\u064f\u0633\u064e\u0645\u064e\u0651\u0627\u0629\u060c \u0628\u062d\u064a\u062b \u064a\u0645\u0643\u0646 \u0644\u0623\u064a \u0628\u0631\u0646\u0627\u0645\u062c \u062a\u0634\u063a\u064a\u0644 \u062a\u062d\u0645\u064a\u0644 \u0623\u064a \u0645\u0644\u0641 GGUF \u062f\u0648\u0646 \u0627\u0644\u062d\u0627\u062c\u0629 \u0625\u0644\u0649 \u0645\u0644\u0641\u0627\u062a \u0645\u0646\u0641\u0635\u0644\u0629 \u0645\u062b\u0644 <code>config.json<\/code> \u0623\u0648 \u0645\u0644\u0641\u0627\u062a \u0627\u0644\u0645\u0639\u0627\u0644\u062c \u0627\u0644\u0644\u063a\u0648\u064a (tokenizer) \u062c\u0646\u0628\u064b\u0627 \u0625\u0644\u0649 \u062c\u0646\u0628 \u0645\u0639\u0647.<\/p>\n<p>\u064a\u0646\u062c\u0645 \u0639\u0646 \u0647\u0630\u0627 \u0627\u0644\u062a\u0635\u0645\u064a\u0645 \u062b\u0644\u0627\u062b\u0629 \u0646\u062a\u0627\u0626\u062c \u0639\u0645\u0644\u064a\u0629:<\/p>\n<ul>\n<li><strong>\u0645\u0644\u0641 \u0648\u0627\u062d\u062f \u064a\u0645\u062b\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u0623\u0643\u0645\u0644\u0647.<\/strong> \u064a\u0645\u0643\u0646\u0643 \u0646\u0633\u062e \u0645\u0644\u0641 <code>.gguf<\/code> \u0628\u064a\u0646 \u0623\u062c\u0647\u0632\u0629 \u0645\u062e\u062a\u0644\u0641\u0629\u060c \u0648\u064a\u064f\u0634\u062a\u063a\u0644 \u0639\u0644\u064a\u0647 \u0645\u0628\u0627\u0634\u0631\u0629\u064b \u062f\u0648\u0646 \u0645\u0634\u0627\u0643\u0644. (\u0623\u0645\u0627 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u062c\u062f\u064b\u0651\u0627 \u0641\u0647\u064a \u062a\u064f\u0642\u0633\u064e\u0651\u0645 \u0623\u062d\u064a\u0627\u0646\u064b\u0627 \u0625\u0644\u0649 \u0623\u062c\u0632\u0627\u0621 \u0645\u0631\u0642\u064e\u0651\u0645\u0629 \u0645\u062b\u0644 <code>-00001-of-00002.gguf<\/code>\u060c \u0644\u0643\u0646\u0647\u0627 \u0645\u0627 \u0632\u0627\u0644\u062a \u062a\u0634\u0643\u0650\u0651\u0644 \u0646\u0645\u0648\u0630\u062c\u064b\u0627 \u0645\u0646\u0637\u0642\u064a\u064b\u0651\u0627 \u0648\u0627\u062d\u062f\u064b\u0627.)<\/li>\n<li><strong>\u064a\u064f\u0645\u0643\u0646 \u062e\u0631\u0652\u0637 \u0627\u0644\u0645\u0644\u0641 \u0641\u064a \u0627\u0644\u0630\u0627\u0643\u0631\u0629 (memory-mappable).<\/strong> \u0648\u062a\u0633\u062a\u0637\u064a\u0639 \u0628\u064a\u0626\u0627\u062a \u0627\u0644\u062a\u0634\u063a\u064a\u0644 (runtimes) <code>\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u062f\u0627\u0644\u0629 \u00abmmap\u00bb<\/code> \u0644\u062e\u0631\u0652\u0637 \u0627\u0644\u0645\u0644\u0641 \u0641\u064a \u0627\u0644\u0630\u0627\u0643\u0631\u0629 \u0648\u062a\u062d\u0645\u064a\u0644 \u0623\u0648\u0632\u0627\u0646\u0647 \u0635\u0641\u062d\u064a\u0629\u064b \u0639\u0646\u062f \u0627\u0644\u062d\u0627\u062c\u0629\u060c \u0645\u0645\u0627 \u064a\u062c\u0639\u0644 \u0639\u0645\u0644\u064a\u0629 \u0627\u0644\u062a\u062d\u0645\u064a\u0644 \u0633\u0631\u064a\u0639\u0629\u064b \u0648\u064a\u0633\u0645\u062d \u0628\u062a\u0634\u063a\u064a\u0644 \u0646\u0645\u0648\u0630\u062c\u064d \u0623\u0643\u0628\u0631 \u0645\u0646 \u0645\u0642\u062f\u0627\u0631 \u0627\u0644\u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0639\u0634\u0648\u0627\u0626\u064a\u0629 (RAM) \u0627\u0644\u0645\u062a\u0627\u062d.<\/li>\n<li><strong>\u0627\u0644\u0643\u0645\u0650\u0651\u064a\u064e\u0651\u0629 (quantisation) \u0645\u062f\u0645\u062c\u0629\u064c \u0641\u064a \u0627\u0644\u062a\u0646\u0633\u064a\u0642.<\/strong> \u064a\u062d\u062a\u0648\u064a \u0645\u0644\u0641 GGUF \u0639\u0644\u0649 \u0627\u0644\u0623\u0648\u0632\u0627\u0646 \u0645\u0636\u063a\u0648\u0637\u0629 \u0645\u0633\u0628\u0642\u064b\u0627 \u0625\u0644\u0649 \u0662\u2013\u0668 \u0628\u062a \u0644\u0643\u0644 \u0648\u0632\u0646\u060c \u0648\u0644\u0630\u0644\u0643 \u0641\u0625\u0646 \u0646\u0645\u0648\u0630\u062c\u064b\u0627 \u0628\u0633\u0639\u0629 \u0668 \u0645\u0644\u064a\u0627\u0631\u0627\u062a \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0642\u062f \u0644\u0627 \u064a\u062a\u062c\u0627\u0648\u0632 \u062d\u062c\u0645\u0647 \u0664,\u0669 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a \u0628\u062f\u0644\u064b\u0627 \u0645\u0646 \u0661\u0666 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a.<\/li>\n<\/ul>\n<h2><span class=\"ez-toc-section\" id=\"Why_GGUF_replaced_GGML\"><\/span>\u0627\u0644\u0633\u0628\u0628 \u0627\u0644\u0630\u064a \u062f\u0641\u0639 \u0625\u0644\u0649 \u0627\u0633\u062a\u0628\u062f\u0627\u0644 GGML \u0628\u0640 GGUF<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0648\u0642\u0628\u0644 \u0634\u0647\u0631 \u0623\u063a\u0633\u0637\u0633 \u0662\u0660\u0662\u0663\u060c \u0643\u0627\u0646 \u0645\u0634\u0631\u0648\u0639 llama.cpp \u064a\u0633\u062a\u062e\u062f\u0645 \u062a\u0646\u0633\u064a\u0642\u064b\u0627 \u064a\u064f\u062f\u0639\u0649 GGML (\u0648\u0647\u0648 \u0627\u0633\u0645\u064c \u0645\u0633\u062a\u0645\u062f\u064c \u0645\u0646 \u0645\u0643\u062a\u0628\u0629 \u0627\u0644\u062a\u0646\u0633\u0648\u0631 \u0627\u0644\u0623\u0633\u0627\u0633\u064a\u0629 \u0627\u0644\u062a\u064a \u064a\u0639\u0645\u0644 \u0639\u0644\u064a\u0647\u0627). \u0648\u0642\u062f \u0643\u0627\u0646 \u062a\u0646\u0633\u064a\u0642 GGML \u0641\u0639\u0651\u0627\u0644\u064b\u0627\u060c \u0644\u0643\u0646\u0647 \u0627\u0634\u062a\u0643\u0649 \u0639\u064a\u0628\u064b\u0627 \u0647\u064a\u0643\u0644\u064a\u064b\u0651\u0627: \u0641\u0642\u062f \u0643\u0627\u0646 \u062a\u0631\u062a\u064a\u0628 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0641\u064a\u0647 \u062c\u0627\u0645\u062f\u064b\u0627 \u0648\u063a\u064a\u0631 \u0645\u064f\u0631\u0642\u064e\u0651\u0645. \u0648\u0643\u0644 \u0645\u0631\u0629 \u0623\u0636\u0627\u0641 \u0641\u064a\u0647\u0627 llama.cpp \u0645\u064a\u0632\u0629 \u062c\u062f\u064a\u062f\u0629 \u2014 \u0645\u062b\u0644 \u0647\u064a\u0627\u0643\u0644 \u0646\u0645\u0627\u0630\u062c \u062c\u062f\u064a\u062f\u0629 \u0623\u0648 \u0637\u0631\u0642 \u0643\u0645\u064a\u0651\u0629 \u0623\u0641\u0636\u0644 \u0623\u0648 \u0645\u0639\u0627\u064a\u064a\u0631 \u0642\u064a\u0627\u0633 \u062a\u062f\u0631\u062c\u0627\u062a \u0627\u0644\u062d\u0628\u0627\u0644 (rope scaling parameters) \u2014 \u0643\u0627\u0646 \u064a\u062a\u0639\u064a\u064e\u0651\u0646 \u062a\u0639\u062f\u064a\u0644 \u0627\u0644\u062a\u0646\u0633\u064a\u0642\u060c \u0645\u0627 \u0623\u062f\u0651\u0649 \u0625\u0644\u0649 \u062a\u0648\u0642\u0641 \u062c\u0645\u064a\u0639 \u0645\u0644\u0641\u0627\u062a \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u0648\u062c\u0648\u062f\u0629 \u0645\u0633\u0628\u0642\u064b\u0627 \u0639\u0644\u0649 \u0643\u0644 \u0627\u0644\u0623\u0642\u0631\u0627\u0635 \u0627\u0644\u0635\u0644\u0628\u0629 \u0639\u0646 \u0627\u0644\u0639\u0645\u0644. \u0648\u0646\u062a\u064a\u062c\u0629\u064b \u0644\u0630\u0644\u0643\u060c \u0627\u0636\u0637\u0631 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u0648\u0646 \u0625\u0644\u0649 \u0625\u0639\u0627\u062f\u0629 \u062a\u0646\u0632\u064a\u0644 \u0623\u0648 \u0625\u0639\u0627\u062f\u0629 \u062a\u062d\u0648\u064a\u0644 \u0645\u062c\u0645\u0648\u0639\u0627\u062a \u0646\u0645\u0627\u0630\u062c\u0647\u0645 \u0628\u0627\u0644\u0643\u0627\u0645\u0644 \u0639\u062f\u0629 \u0645\u0631\u0627\u062a \u062e\u0644\u0627\u0644 \u0628\u0636\u0639\u0629 \u0623\u0634\u0647\u0631.<\/p>\n<p>\u0623\u0645\u0651\u0627 \u062a\u0646\u0633\u064a\u0642 GGUF\u060c \u0627\u0644\u0630\u064a \u0642\u062f\u0651\u0645\u0647 \u0645\u0634\u0631\u0648\u0639 llama.cpp \u0641\u064a \u0623\u063a\u0633\u0637\u0633 \u0662\u0660\u0662\u0663\u060c \u0641\u0642\u062f \u0639\u0627\u0644\u062c \u0647\u0630\u0647 \u0627\u0644\u0645\u0634\u0643\u0644\u0629 \u0639\u0628\u0631 \u062a\u063a\u064a\u064a\u0631\u064a\u0646:<\/p>\n<ul>\n<li><strong>\u0627\u0644\u062a\u0642\u0633\u064a\u0645 \u0625\u0644\u0649 \u0625\u0635\u062f\u0627\u0631\u0627\u062a (versioning).<\/strong> \u0641\u064a\u0639\u0644\u0646 \u0627\u0644\u0645\u0644\u0641 \u0639\u0646 \u0625\u0635\u062f\u0627\u0631 \u062a\u0646\u0633\u064a\u0642\u0647\u060c \u0644\u0643\u064a \u064a\u0639\u0631\u0641 \u0642\u0627\u0631\u0626\u0648 \u0627\u0644\u0645\u0644\u0641 \u0628\u0627\u0644\u0636\u0628\u0637 \u0643\u064a\u0641\u064a\u0629 \u062a\u062d\u0644\u064a\u0644\u0647.<\/li>\n<li><strong>\u0628\u064a\u0627\u0646\u0627\u062a \u0648\u0635\u0641\u064a\u0629 \u0642\u0627\u0628\u0644\u0629 \u0644\u0644\u062a\u0648\u0633\u0651\u0639 \u0639\u0644\u0649 \u0634\u0643\u0644 \u0623\u0632\u0648\u0627\u062c \u0645\u0641\u062a\u0627\u062d-\u0642\u064a\u0645\u0629 (extensible key-value metadata).<\/strong> \u0623\u0645\u0627 \u0627\u0644\u0645\u0639\u0644\u0648\u0645\u0627\u062a \u0627\u0644\u062c\u062f\u064a\u062f\u0629 (\u0645\u062b\u0644 \u062d\u0642\u0644 \u062c\u062f\u064a\u062f \u0644\u0647\u064a\u0643\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0623\u0648 \u0642\u0627\u0644\u0628 \u062f\u0631\u062f\u0634\u0629 \u0623\u0648 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0641\u0627\u0626\u0642\u0629 \u0625\u0636\u0627\u0641\u064a\u0629) \u0641\u0647\u064a \u0645\u062c\u0631\u0651\u062f \u0645\u0641\u062a\u0627\u062d \u062c\u062f\u064a\u062f. \u0648\u0628\u0630\u0644\u0643 \u062a\u0638\u0644 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0627\u0644\u0642\u062f\u064a\u0645\u0629 \u0627\u0644\u062a\u064a \u0644\u0627 \u062a\u062d\u062a\u0648\u064a \u0630\u0644\u0643 \u0627\u0644\u0645\u0641\u062a\u0627\u062d \u0642\u0627\u0628\u0644\u0629 \u0644\u0644\u062a\u062d\u0645\u064a\u0644\u060c \u0628\u064a\u0646\u0645\u0627 \u062a\u0639\u0645\u0644 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0627\u0644\u062c\u062f\u064a\u062f\u0629 \u0627\u0644\u062a\u064a \u062a\u062a\u0636\u0645\u0651\u0646\u0647 \u0641\u064a \u0628\u064a\u0626\u0627\u062a \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0645\u062d\u062f\u064e\u0651\u062b\u0629. \u0648\u0644\u0627 \u064a\u062a\u0633\u0628\u0651\u0628 \u0623\u064a \u062a\u063a\u064a\u064a\u0631 \u0641\u064a \u0643\u0633\u0631 \u0627\u0644\u062a\u0648\u0627\u0641\u0642 \u0627\u0644\u0639\u0643\u0633\u064a (retroactive breakage).<\/li>\n<\/ul>\n<p>\u0648\u0642\u062f \u0623\u0644\u063a\u0649 llama.cpp \u062f\u0639\u0645 GGML \u0628\u0639\u062f \u0641\u062a\u0631\u0629 \u0648\u062c\u064a\u0632\u0629\u060c \u0648\u062a\u0628\u0639\u0647 \u0627\u0644\u0646\u0638\u0627\u0645 \u0627\u0644\u0628\u064a\u0626\u064a \u0643\u0644\u0647. \u0648\u0627\u0644\u064a\u0648\u0645 \u0644\u0645 \u064a\u0628\u0642\u064e GGML \u0633\u0648\u0649 \u0643\u0627\u0633\u0645\u064d \u0644\u0645\u0643\u062a\u0628\u0629 \u0627\u0644\u062a\u0646\u0633\u0648\u0631 \u0627\u0644\u0623\u0633\u0627\u0633\u064a\u0629\u061b \u0641\u0625\u0630\u0627 \u0648\u062c\u062f\u062a \u0645\u0644\u0641 \u0646\u0645\u0648\u0630\u062c \u0630\u0627 \u0627\u0644\u0627\u0645\u062a\u062f\u0627\u062f <code>.ggml<\/code> \u0623\u0648 <code>\u0623\u0648 .bin<\/code> \u0645\u0646 \u0639\u0627\u0645 \u0662\u0660\u0662\u0663\u060c \u0641\u0644\u0646 \u064a\u064f\u062d\u0645\u064e\u0651\u0644 \u0641\u064a \u0623\u064a \u0623\u062f\u0627\u0629 \u062d\u0627\u0644\u064a\u0629\u060c \u0628\u0644 \u064a\u0646\u0628\u063a\u064a \u0639\u0644\u064a\u0643 \u062a\u0646\u0632\u064a\u0644 \u0646\u0633\u062e\u0629 \u0645\u0646\u0647 \u0628\u062a\u0646\u0633\u064a\u0642 GGUF \u0628\u062f\u0644\u064b\u0627 \u0645\u0646 \u0630\u0644\u0643.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"How_to_read_the_quantisation_suffixes\"><\/span>\u0643\u064a\u0641\u064a\u0629 \u0642\u0631\u0627\u0621\u0629 \u0644\u0648\u0627\u062d\u0642 \u0627\u0644\u062a\u0643\u0645\u064a\u0645<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u064a\u062d\u0645\u0644 \u0643\u0644 \u0627\u0633\u0645 \u0645\u0644\u0641 GGUF \u0644\u0627\u062d\u0642\u0629\u064b \u0645\u062b\u0644 <code>Q4_K_M<\/code> \u062a\u0648\u0636\u062d \u0645\u062f\u0649 \u0634\u062f\u0629 \u0636\u063a\u0637 \u0627\u0644\u0623\u0648\u0632\u0627\u0646. \u0648\u0627\u0644\u0646\u0645\u0637 \u0647\u0648:<\/p>\n<ul>\n<li><strong>\u0641\u0627\u0644\u0631\u0642\u0645 \u0627\u0644\u0648\u0627\u0642\u0639 \u0628\u0639\u062f \u062d\u0631\u0641 Q<\/strong> \u064a\u0634\u064a\u0631 \u062a\u0642\u0631\u064a\u0628\u064b\u0627 \u0625\u0644\u0649 \u0639\u062f\u062f \u0627\u0644\u0628\u062a\u0627\u062a \u0644\u0643\u0644 \u0648\u0632\u0646: \u0641\u0640 Q4 \u062a\u0639\u0646\u064a \u062a\u0642\u0631\u064a\u0628\u064b\u0627 \u0664 \u0628\u062a\u060c \u0648Q8 \u062a\u0639\u0646\u064a \u062a\u0642\u0631\u064a\u0628\u064b\u0627 \u0668 \u0628\u062a. \u0648\u0643\u0644\u0645\u0627 \u0642\u0644\u0651 \u0639\u062f\u062f \u0627\u0644\u0628\u062a\u0627\u062a\u060c \u0642\u0644\u0651 \u062d\u062c\u0645 \u0627\u0644\u0645\u0644\u0641 \u0648\u0627\u0646\u062e\u0641\u0636\u062a \u062c\u0648\u062f\u062a\u0647.<\/li>\n<li><strong>_K<\/strong> \u062a\u0634\u064a\u0631 \u0625\u0644\u0649 \u0637\u0631\u0642 \u0627\u0644\u0643\u0645\u0651 \u0627\u0644\u062c\u062f\u064a\u062f\u0629 \u0627\u0644\u0645\u064f\u0633\u0645\u0627\u0629 \u00abk-quant\u00bb\u060c \u0648\u0627\u0644\u062a\u064a \u062a\u062e\u0635\u0635 \u0628\u062a\u0627\u062a \u0625\u0636\u0627\u0641\u064a\u0629 \u0644\u0644\u062a\u064a\u0646\u0633\u0648\u0631\u0627\u062a \u0627\u0644\u0623\u0643\u062b\u0631 \u062a\u0623\u062b\u064a\u0631\u064b\u0627 \u0641\u064a \u062c\u0648\u062f\u0629 \u0627\u0644\u0645\u062e\u0631\u062c\u0627\u062a. \u0648\u0628\u0627\u0644\u0645\u0642\u0627\u0631\u0646\u0629 \u0639\u0646\u062f \u0646\u0641\u0633 \u0627\u0644\u062d\u062c\u0645\u060c \u0641\u0625\u0646 \u0627\u0644\u0643\u0645\u0651 \u0645\u0646 \u0646\u0648\u0639 K \u064a\u062a\u0641\u0648\u0651\u0642 \u0639\u0644\u0649 \u0627\u0644\u0637\u0631\u064a\u0642\u0629 \u0627\u0644\u0623\u0642\u062f\u0645.<\/li>\n<li><strong>_S \/ _M \/ _L<\/strong> (\u0635\u063a\u064a\u0631\/\u0645\u062a\u0648\u0633\u0637\/\u0643\u0628\u064a\u0631) \u0647\u064a \u0645\u062a\u063a\u064a\u0631\u0627\u062a \u0636\u0645\u0646 \u0639\u0627\u0626\u0644\u0629 K \u2014 \u062d\u064a\u062b \u0625\u0646 <code>_M<\/code> \u062a\u062d\u062a\u0641\u0638 \u0628\u0639\u062f\u062f \u0642\u0644\u064a\u0644 \u0645\u0646 \u0627\u0644\u062a\u064a\u0646\u0633\u0648\u0631\u0627\u062a \u0627\u0644\u062d\u0633\u0627\u0633\u0629 \u0628\u062f\u0642\u0629 \u0623\u0639\u0644\u0649 \u0645\u0646 \u062a\u0644\u0643 \u0627\u0644\u062a\u064a \u062a\u062d\u062a\u0641\u0638 \u0628\u0647\u0627 <code>_S<\/code>.<\/li>\n<li><strong>_0<\/strong> (\u0623\u0645\u0627 <code>Q8_0<\/code>, <code>Q4_0<\/code>\u0641\u062a\u0634\u064a\u0631 \u0625\u0644\u0649 \u0637\u0631\u064a\u0642\u0629 \u0627\u0644\u0643\u0645\u0651 \u0627\u0644\u0643\u062a\u0644\u064a \u0627\u0644\u0628\u0633\u064a\u0637\u0629 \u0627\u0644\u0623\u0642\u062f\u0645.) <code>Q8_0<\/code> \u0648\u0644\u0627 \u062a\u0632\u0627\u0644 \u062a\u064f\u0633\u062a\u062e\u062f\u0645 \u0639\u0644\u0649 \u0646\u0637\u0627\u0642 \u0648\u0627\u0633\u0639 \u0644\u0623\u0646 \u0627\u0644\u0637\u0631\u064a\u0642\u0629 \u0627\u0644\u0628\u0633\u064a\u0637\u0629 \u0639\u0646\u062f \u0668 \u0628\u062a \u062a\u0643\u0648\u0646 \u0642\u0631\u064a\u0628\u0629 \u062c\u062f\u064b\u0651\u0627 \u0645\u0646 \u062d\u0627\u0644\u0629 \u0639\u062f\u0645 \u0627\u0644\u0641\u0642\u062f\u0627\u0646 (near-lossless)\u060c \u0623\u0645\u0627 <code>Q4_0<\/code> \u0641\u0642\u062f \u0623\u0635\u0628\u062d\u062a \u0642\u062f\u064a\u0645\u0629\u064b \u0625\u0644\u0649 \u062d\u062f\u064d\u0651 \u0643\u0628\u064a\u0631 \u2014 \u0644\u0630\u0627 \u064a\u064f\u0641\u0636\u064e\u0651\u0644 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 <code>Q4_K_M<\/code>.<\/li>\n<li><strong>\u0628\u0627\u062f\u0626\u0627\u062a IQ<\/strong> (\u0645\u062b\u0644<code>IQ2_M<\/code>, <code>\u0648IQ3_XS<\/code>\u0625\u0644\u062e) \u0648\u0647\u064a \u0637\u0631\u0642 \u00ab\u0643\u0645\u0651 \u0623\u0647\u0645\u064a\u00bb (i-quants) \u062a\u064f\u0628\u0646\u0649 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0645\u0635\u0641\u0648\u0641\u0629 \u0623\u0647\u0645\u064a\u0629\u060c \u0648\u062a\u064f\u0633\u062a\u062e\u062f\u064e\u0645 \u0644\u0636\u063a\u0637 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0625\u0644\u0649 \u0623\u0642\u0644 \u0645\u0646 \u0664 \u0628\u062a \u0645\u0639 \u062a\u0642\u0644\u064a\u0644 \u0627\u0644\u0636\u0631\u0631 \u0645\u0642\u0627\u0631\u0646\u0629\u064b \u0628\u0627\u0644\u0628\u062f\u0627\u0626\u0644 \u0627\u0644\u0623\u062e\u0631\u0649.<\/li>\n<li><strong>F16 \/ BF16 \/ F32<\/strong> \u062a\u0639\u0646\u064a \u0623\u0646 \u0627\u0644\u0623\u0648\u0632\u0627\u0646 \u063a\u064a\u0631 \u0645\u0643\u0645\u064e\u0651\u0645\u0629 (unquantised) \u0648\u0628\u062f\u0642\u0629 \u0661\u0666 \u0623\u0648 \u0663\u0662 \u0628\u062a \u2014 \u0648\u0647\u064a \u0627\u0644\u062c\u0648\u062f\u0629 \u0627\u0644\u0645\u0631\u062c\u0639\u064a\u0629\u060c \u0648\u0623\u0643\u0628\u0631 \u062d\u062c\u0645 \u0645\u0645\u0643\u0646.<\/li>\n<\/ul>\n<p>\u0648\u0641\u064a\u0645\u0627 \u064a\u0644\u064a \u062a\u0643\u0644\u0641\u0629 \u0627\u0644\u062e\u064a\u0627\u0631\u0627\u062a \u0627\u0644\u0634\u0627\u0626\u0639\u0629 \u0639\u0645\u0644\u064a\u064b\u0651\u0627\u060c \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0646\u0645\u0627\u0630\u062c Llama \u0645\u0646 \u0641\u0626\u0629 \u0668 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 (instruct models) \u0643\u0645\u062b\u0627\u0644 (\u0645\u0639 \u0627\u0644\u0639\u0644\u0645 \u0623\u0646 \u0627\u0644\u0623\u062d\u062c\u0627\u0645 \u062a\u0642\u0631\u064a\u0628\u064a\u0629 \u0648\u062a\u062a\u0641\u0627\u0648\u062a \u0642\u0644\u064a\u0644\u064b\u0627 \u0628\u0627\u062e\u062a\u0644\u0627\u0641 \u0627\u0644\u0646\u0645\u0648\u0630\u062c):<\/p>\n<table>\n<thead>\n<tr>\n<th>\u0627\u0644\u0643\u0645\u0651<\/th>\n<th>\u062d\u062c\u0645 \u0627\u0644\u0645\u0644\u0641 (\u0646\u0645\u0648\u0630\u062c \u0668 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629)<\/th>\n<th>\u0627\u0644\u062c\u0648\u062f\u0629 \u0645\u0642\u0627\u0631\u0646\u0629\u064b \u0628\u0640 F16<\/th>\n<th>\u0645\u062a\u0649 \u062a\u064f\u0633\u062a\u062e\u062f\u0645<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>F16<\/td>\n<td>~\u0661\u0666,\u0661 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0645\u0631\u062c\u0639\u064a\u0629<\/td>\n<td>\u0644\u0623\u063a\u0631\u0627\u0636 \u0627\u0644\u0645\u0642\u0627\u0631\u0646\u0629 \u0627\u0644\u062a\u062c\u0631\u064a\u0628\u064a\u0629 (benchmarking) \u0623\u0648 \u0644\u0625\u062c\u0631\u0627\u0621 \u0643\u0645\u0651 \u0625\u0636\u0627\u0641\u064a\u061b \u0648\u0646\u0627\u062f\u0631\u064b\u0627 \u0645\u0627 \u062a\u0633\u062a\u062d\u0642 \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0645\u0628\u0627\u0634\u0631<\/td>\n<\/tr>\n<tr>\n<td>Q8_0<\/td>\n<td>~\u0668,\u0665 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0644\u0627 \u064a\u0645\u0643\u0646 \u062a\u0645\u064a\u064a\u0632 \u0627\u0644\u0641\u0631\u0642 \u0639\u0645\u0644\u064a\u064b\u0651\u0627<\/td>\n<td>\u0625\u0630\u0627 \u0643\u0627\u0646\u062a \u0644\u062f\u064a\u0643 \u0630\u0627\u0643\u0631\u0629 \u0641\u064a\u062f\u064a\u0648 (VRAM) \u0648\u0627\u0641\u0631\u0629 \u0648\u062a\u0631\u064a\u062f \u0623\u0642\u0635\u0649 \u062c\u0648\u062f\u0629 \u0645\u0645\u0643\u0646\u0629<\/td>\n<\/tr>\n<tr>\n<td>Q6_K<\/td>\n<td>~\u0666,\u0666 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0641\u0642\u062f\u0627\u0646 \u0636\u0626\u064a\u0644 \u062c\u062f\u064b\u0651\u0627<\/td>\n<td>\u062c\u0648\u062f\u0629 \u0645\u0645\u062a\u0627\u0632\u0629 \u0645\u0639 \u062a\u0648\u0641\u064a\u0631 \u0645\u0644\u062d\u0648\u0638 \u0641\u064a \u0627\u0644\u062d\u062c\u0645<\/td>\n<\/tr>\n<tr>\n<td>Q5_K_M<\/td>\n<td>~5.7 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0641\u0642\u062f\u0627\u0646 \u0637\u0641\u064a\u0641 \u062c\u062f\u064b\u0651\u0627<\/td>\n<td>\u062a\u0648\u0627\u0632\u0646 \u062c\u064a\u062f \u0628\u064a\u0646 \u0627\u0644\u062c\u0648\u062f\u0629 \u0648\u0627\u0644\u062d\u062c\u0645<\/td>\n<\/tr>\n<tr>\n<td>Q4_K_M<\/td>\n<td>~4.9 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0641\u0642\u062f\u0627\u0646 \u0628\u0633\u064a\u0637\u060c \u0648\u063a\u0627\u0644\u0628\u064b\u0627 \u0645\u0627 \u064a\u0643\u0648\u0646 \u0645\u0642\u0628\u0648\u0644\u064b\u0627<\/td>\n<td><strong>\u0627\u0644\u0642\u064a\u0645\u0629 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a\u0629 \u0627\u0644\u0642\u064a\u0627\u0633\u064a\u0629.<\/strong> \u0623\u0641\u0636\u0644 \u0646\u0633\u0628\u0629 \u062c\u0648\u062f\u0629 \u0625\u0644\u0649 \u0627\u0644\u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a \u0644\u0645\u0639\u0638\u0645 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646<\/td>\n<\/tr>\n<tr>\n<td>Q3_K_M<\/td>\n<td>~4.0 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u062a\u062f\u0647\u0648\u0631 \u0645\u0644\u062d\u0648\u0638 \u0641\u064a \u0627\u0644\u062c\u0648\u062f\u0629<\/td>\n<td>\u0641\u0642\u0637 \u0639\u0646\u062f\u0645\u0627 \u0644\u0627 \u064a\u062a\u0646\u0627\u0633\u0628 \u0627\u0644\u062a\u0643\u0645\u064a\u0645 Q4 \u0641\u0639\u0644\u064a\u064b\u0651\u0627 \u0645\u0639 \u0627\u0644\u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0645\u062a\u0648\u0641\u0631\u0629<\/td>\n<\/tr>\n<tr>\n<td>Q2_K \/ IQ2<\/td>\n<td>~3.2 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u062a\u062f\u0647\u0648\u0631 \u0643\u0628\u064a\u0631 \u0641\u064a \u0627\u0644\u062c\u0648\u062f\u0629<\/td>\n<td>\u0645\u0644\u0627\u0630 \u0623\u062e\u064a\u0631\u061b \u0625\u0630 \u0625\u0646 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0646\u0645\u0648\u0630\u062c \u0623\u0635\u063a\u0631 \u0628\u0643\u0641\u0627\u0621\u0629 Q4 \u063a\u0627\u0644\u0628\u064b\u0627 \u0645\u0627 \u064a\u0643\u0648\u0646 \u0623\u0641\u0636\u0644<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u0642\u0648\u0627\u0639\u0650\u062f\u064f \u0645\u0641\u064a\u062f\u0629\u064c \u0627\u062b\u0646\u062a\u0627\u0646: \u064a\u0624\u062b\u0631 \u0627\u0644\u062a\u0643\u0645\u064a\u0645 \u0633\u0644\u0628\u064b\u0627 \u0639\u0644\u0649 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0635\u063a\u064a\u0631\u0629 \u0623\u0643\u062b\u0631 \u0645\u0646 \u062a\u0623\u062b\u064a\u0631\u0647 \u0639\u0644\u0649 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0643\u0628\u064a\u0631\u0629 (\u0641\u0646\u0645\u0648\u0630\u062c 70 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q3 \u064a\u062d\u062a\u0641\u0638 \u0628\u062c\u0648\u062f\u062a\u0647 \u0628\u0634\u0643\u0644\u064d \u0623\u0641\u0636\u0644 \u0628\u0643\u062b\u064a\u0631\u064d \u0645\u0646 \u0646\u0645\u0648\u0630\u062c 8 \u0645\u0644\u064a\u0627\u0631\u0627\u062a \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0646\u0641\u0633 \u0627\u0644\u0645\u0633\u062a\u0648\u0649 Q3)\u060c \u0643\u0645\u0627 \u0623\u0646 \u0645\u0646\u062d\u0646\u0649 \u0627\u0644\u062c\u0648\u062f\u0629 \u064a\u0646\u062e\u0641\u0636 \u0627\u0646\u062e\u0641\u0627\u0636\u064b\u0627 \u062d\u0627\u062f\u064b\u0651\u0627 \u062a\u062d\u062a \u0645\u0633\u062a\u0648\u0649 Q4. \u0648\u0639\u0646\u062f \u0627\u0644\u0634\u0643\u060c <code>Q4_K_M<\/code> \u0647\u0648 \u0627\u0644\u062e\u064a\u0627\u0631 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a \u0644\u0644\u0645\u062c\u062a\u0645\u0639 \u0648\u0644\u0633\u0628\u0628 \u0648\u062c\u064a\u0647.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Picking_a_quantisation_for_your_VRAM\"><\/span>\u0627\u062e\u062a\u064a\u0627\u0631 \u062f\u0631\u062c\u0629 \u062a\u0643\u0645\u064a\u0645 \u0645\u0646\u0627\u0633\u0628\u0629 \u0644\u0633\u0639\u0629 \u0630\u0627\u0643\u0631\u0629 VRAM \u0644\u062f\u064a\u0643<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u0625\u062c\u0645\u0627\u0644\u064a \u0627\u0644\u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0645\u0637\u0644\u0648\u0628\u0629 \u064a\u0633\u0627\u0648\u064a \u062a\u0642\u0631\u064a\u0628\u064b\u0627 <em>\u062d\u062c\u0645 \u0627\u0644\u0645\u0644\u0641 + \u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u062a\u062e\u0632\u064a\u0646 \u0627\u0644\u0645\u0624\u0642\u062a \u0644\u0644\u0633\u064a\u0627\u0642 + \u0627\u0644\u0647\u0648\u0627\u0645\u0634 \u0627\u0644\u0625\u0636\u0627\u0641\u064a\u0629<\/em>. \u062e\u0635\u0650\u0651\u0635 \u0645\u064a\u0632\u0627\u0646\u064a\u0629 \u0625\u0636\u0627\u0641\u064a\u0629 \u0642\u062f\u0631\u0647\u0627 1\u20132 \u063a\u064a\u063a\u0627\u0628\u0627\u064a\u062a \u0641\u0648\u0642 \u062d\u062c\u0645 \u0627\u0644\u0645\u0644\u0641 \u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0628\u0636\u0639\u0629 \u0622\u0644\u0627\u0641 \u0645\u0646 \u0627\u0644\u0631\u0645\u0648\u0632 (tokens) \u0636\u0645\u0646 \u0627\u0644\u0633\u064a\u0627\u0642\u060c \u0648\u0632\u0650\u062f \u0647\u0630\u0647 \u0627\u0644\u0645\u064a\u0632\u0627\u0646\u064a\u0629 \u0625\u0630\u0627 \u0643\u0646\u062a \u062a\u0633\u062a\u062e\u062f\u0645 \u0633\u064a\u0627\u0642\u0627\u062a \u0637\u0648\u064a\u0644\u0629 \u062c\u062f\u064b\u0651\u0627. \u0648\u064a\u0624\u062f\u064a \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u0623\u0642\u0635\u0649 \u0633\u0631\u0639\u0629\u064d \u0645\u0645\u0643\u0646\u0629\u064d \u0639\u0646\u062f\u0645\u0627 \u062a\u062a\u0633\u0639 \u0643\u0644 \u0647\u0630\u0647 \u0627\u0644\u0645\u0643\u0648\u0646\u0627\u062a \u062f\u0627\u062e\u0644 \u0630\u0627\u0643\u0631\u0629 VRAM \u0627\u0644\u062e\u0627\u0635\u0629 \u0628\u0627\u0644\u0648\u062d\u062f\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u064a\u0629 (GPU)\u061b \u0623\u0645\u0627 \u0623\u062f\u0648\u0627\u062a llama.cpp \u0641\u0647\u064a \u0642\u0627\u062f\u0631\u0629 \u0639\u0644\u0649 \u062a\u062d\u0645\u064a\u0644 \u0627\u0644\u0623\u062c\u0632\u0627\u0621 \u0627\u0644\u0645\u062a\u0628\u0642\u064a\u0629 \u0625\u0644\u0649 \u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0646\u0638\u0627\u0645 \u0627\u0644\u0631\u0626\u064a\u0633\u064a\u0629 (RAM)\u060c \u0648\u0647\u064a \u062a\u0639\u0645\u0644 \u0644\u0643\u0646\u0647\u0627 \u062a\u062a\u0628\u0627\u0637\u0623 \u0643\u062b\u064a\u0631\u064b\u0627 \u0643\u0644\u0645\u0627 \u0632\u0627\u062f \u0639\u062f\u062f \u0627\u0644\u0637\u0628\u0642\u0627\u062a \u0627\u0644\u062a\u064a \u062a\u064f\u062d\u0645\u064e\u0651\u0644 \u062e\u0627\u0631\u062c \u0627\u0644\u0640 VRAM.<\/p>\n<table>\n<thead>\n<tr>\n<th>\u0630\u0627\u0643\u0631\u0629 VRAM<\/th>\n<th>\u0645\u0627 \u064a\u062a\u0646\u0627\u0633\u0628 \u0628\u0633\u0647\u0648\u0644\u0629<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>8 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0646\u0645\u0627\u0630\u062c 7\u20138 \u0645\u0644\u064a\u0627\u0631\u0627\u062a \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q4_K_M \u0623\u0648 Q5_K_M<\/td>\n<\/tr>\n<tr>\n<td>12 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0646\u0645\u0648\u0630\u062c 8 \u0645\u0644\u064a\u0627\u0631\u0627\u062a \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q8_0\u060c \u0623\u0648 \u0646\u0645\u0648\u0630\u062c 12\u201314 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q4_K_M<\/td>\n<\/tr>\n<tr>\n<td>16 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0646\u0645\u0648\u0630\u062c 14 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q5_K_M\/Q6_K<\/td>\n<\/tr>\n<tr>\n<td>24 \u062c\u064a\u062c\u0627\u0628\u0627\u064a\u062a<\/td>\n<td>\u0646\u0645\u0648\u0630\u062c ~32 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q4_K_M\u060c \u0623\u0648 \u0646\u0645\u0648\u0630\u062c 14 \u0645\u0644\u064a\u0627\u0631 \u0645\u0639\u0644\u064e\u0651\u0645\u0629 \u0639\u0646\u062f \u0645\u0633\u062a\u0648\u0649 Q8_0 \u0645\u0639 \u0633\u064a\u0627\u0642 \u0637\u0648\u064a\u0644<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u0644\u0644\u062d\u0635\u0648\u0644 \u0639\u0644\u0649 \u0623\u0631\u0642\u0627\u0645 \u062f\u0642\u064a\u0642\u0629 \u0644\u0645\u064f\u062e\u0637\u064e\u0651\u0637 \u0645\u0639\u064a\u0646 \u0648\u0637\u0648\u0644 \u0633\u064a\u0627\u0642 \u0645\u0639\u064a\u064e\u0651\u0646\u060c \u0641\u0625\u0646 <a href=\"https:\/\/convly.ai\/ar\/llm-vram-calculator\/\">\u062d\u0627\u0633\u0628\u0629 \u0627\u0644\u0630\u0627\u0643\u0631\u0629 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a\u0629 (VRAM)<\/a> \u064a\u0642\u0648\u0645 \u062a\u0644\u0642\u0627\u0626\u064a\u064b\u0651\u0627 \u0628\u0647\u0630\u0647 \u0627\u0644\u062d\u0633\u0627\u0628\u0627\u062a \u0646\u064a\u0627\u0628\u0629\u064b \u0639\u0646\u0643\u060c \u0648\u0647\u0646\u0627\u0643 \u062a\u0641\u0635\u064a\u0644\u064c \u0644\u0643\u0644 \u0646\u0645\u0648\u0630\u062c \u0641\u064a <a href=\"https:\/\/convly.ai\/ar\/vram-requirements-every-major-llm-2026\/\">\u0645\u062a\u0637\u0644\u0628\u0627\u062a \u0630\u0627\u0643\u0631\u0629 VRAM \u0644\u0643\u0644 \u0646\u0645\u0648\u0630\u062c \u0644\u063a\u0648\u064a \u0631\u0626\u064a\u0633\u064a<\/a>. \u0648\u0625\u0630\u0627 \u0643\u0646\u062a \u062a\u062e\u062a\u0627\u0631 \u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0628\u062f\u0644\u064b\u0627 \u0645\u0646 \u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062a\u0643\u0645\u064a\u0645\u060c \u0641\u0627\u0628\u062f\u0623 \u0628\u0640 <a href=\"https:\/\/convly.ai\/ar\/best-gpus-for-local-llms-2026\/\">\u0623\u0641\u0636\u0644 \u0648\u062d\u062f\u0627\u062a \u0645\u0639\u0627\u0644\u062c\u0629 \u0631\u0633\u0648\u0645\u064a\u0629 (GPUs) \u0644\u0623\u0646\u0638\u0645\u0629 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0644\u063a\u0648\u064a\u0629 \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u0627\u0644\u0645\u062d\u0644\u064a\u0629 (local LLMs)<\/a>.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"GGUF_vs_safetensors\"><\/span>\u0645\u0642\u0627\u0631\u0646\u0629 \u0628\u064a\u0646 GGUF \u0648safetensors<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u064a\u062a\u0639\u0627\u064a\u0634 \u0647\u0630\u0627\u0646 \u0627\u0644\u062a\u0646\u0633\u064a\u0642\u0627\u0646 \u0645\u0639\u064b\u0627 \u0644\u0623\u0646 \u0643\u0644\u064b\u0651\u0627 \u0645\u0646\u0647\u0645\u0627 \u064a\u062e\u062f\u0645 \u0628\u064a\u0626\u0627\u062a \u062a\u0634\u063a\u064a\u0644 \u0645\u062e\u062a\u0644\u0641\u0629\u060c \u0648\u0644\u064a\u0633 \u0644\u0623\u0646 \u0623\u062d\u062f\u0647\u0645\u0627 \u064a\u062a\u0641\u0648\u0642 \u0639\u0644\u0649 \u0627\u0644\u0622\u062e\u0631.<\/p>\n<table>\n<thead>\n<tr>\n<th><\/th>\n<th>GGUF<\/th>\n<th>safetensors<\/th>\n<\/tr>\n<\/thead>\n<tbody>\n<tr>\n<td>\u0645\u064f\u0635\u0645\u064e\u0651\u0645 \u062e\u0635\u064a\u0635\u064b\u0627 \u0644\u0640<\/td>\n<td>llama.cpp \u0648\u0628\u064a\u0626\u062a\u0647 \u0627\u0644\u062a\u0634\u063a\u064a\u0644\u064a\u0629<\/td>\n<td>\u0628\u064a\u0626\u0629 Hugging Face \/ PyTorch<\/td>\n<\/tr>\n<tr>\n<td>\u0627\u0644\u0645\u062d\u062a\u0648\u064a\u0627\u062a<\/td>\n<td>\u0627\u0644\u0623\u0648\u0632\u0627\u0646 + \u0645\u0634\u0641\u0631 \u0627\u0644\u0631\u0645\u0648\u0632 (tokenizer) + \u0645\u0644\u0641 \u0627\u0644\u062a\u0647\u064a\u0626\u0629 (config) \u0641\u064a \u0645\u0644\u0641 \u0648\u0627\u062d\u062f<\/td>\n<td>\u0627\u0644\u0642\u064a\u0645 \u0627\u0644\u0639\u062f\u062f\u064a\u0629 \u0641\u0642\u0637 (tensors)\u061b \u0628\u064a\u0646\u0645\u0627 \u064a\u064f\u062e\u0632\u064e\u0651\u0646 \u0645\u0644\u0641 \u0627\u0644\u062a\u0647\u064a\u0626\u0629 \u0648\u0645\u0634\u0641\u0631 \u0627\u0644\u0631\u0645\u0648\u0632 \u0641\u064a \u0645\u0644\u0641\u0627\u062a JSON \u0645\u0646\u0641\u0635\u0644\u0629<\/td>\n<\/tr>\n<tr>\n<td>\u0627\u0644\u062a\u0643\u0645\u064a\u0645 (Quantisation)<\/td>\n<td>\u0645\u062f\u0645\u062c\u064c \u0641\u064a \u0627\u0644\u062a\u0646\u0633\u064a\u0642 (\u0645\u062b\u0644 Q4_K_M \u0625\u0644\u062e.)<\/td>\n<td>\u0639\u0627\u062f\u0629\u064b \u0645\u0627 \u062a\u0643\u0648\u0646 \u0628\u0635\u064a\u063a\u0629 FP16\/BF16\u061b \u0648\u064a\u062a\u0645 \u062a\u0643\u0645\u064a\u0645\u0647\u0627 \u0639\u0628\u0631 \u0623\u0633\u0627\u0644\u064a\u0628 \u0645\u0646\u0641\u0635\u0644\u0629 (\u0645\u062b\u0644 GPTQ \u0648AWQ)<\/td>\n<\/tr>\n<tr>\n<td>\u0628\u064a\u0626\u0627\u062a \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c\u064a\u0629<\/td>\n<td>Ollama\u060c LM Studio\u060c llama.cpp\u060c KoboldCpp\u060c Jan<\/td>\n<td>transformers\u060c vLLM\u060c TGI\u060c SGLang<\/td>\n<\/tr>\n<tr>\n<td>\u0627\u0644\u0646\u0637\u0627\u0642 \u0627\u0644\u0623\u0645\u062b\u0644<\/td>\n<td>\u0627\u0644\u0623\u062c\u0647\u0632\u0629 \u0627\u0644\u0627\u0633\u062a\u0647\u0644\u0627\u0643\u064a\u0629\u060c \u0648\u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u0647\u062c\u064a\u0646 \u0628\u064a\u0646 \u0648\u062d\u062f\u0629 \u0627\u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0645\u0631\u0643\u0632\u064a\u0629 (CPU) \u0648\u0627\u0644\u0648\u062d\u062f\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u064a\u0629 (GPU)\u060c \u0648\u0627\u0644\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0641\u0631\u062f\u064a<\/td>\n<td>\u0648\u062d\u062f\u0627\u062a \u0645\u0639\u0627\u0644\u062c\u0629 \u0631\u0633\u0648\u0645\u064a\u0629 (GPUs) \u0645\u062e\u0635\u0635\u0629 \u0644\u0645\u0631\u0627\u0643\u0632 \u0627\u0644\u0628\u064a\u0627\u0646\u0627\u062a\u060c \u0648\u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0639\u0627\u0644\u064a \u0627\u0644\u0625\u0646\u062a\u0627\u062c\u064a\u0629\u060c \u0648\u0627\u0644\u062a\u062f\u0631\u064a\u0628<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<p>\u0627\u0644\u0642\u0631\u0627\u0631 \u064a\u062a\u0639\u0644\u064e\u0651\u0642 \u0641\u064a \u0627\u0644\u0648\u0627\u0642\u0639 \u0628\u0627\u0644\u0628\u0631\u0645\u062c\u064a\u0627\u062a \u0627\u0644\u062a\u064a \u062a\u0633\u062a\u062e\u062f\u0645\u0647\u0627: \u0641\u0628\u0631\u0627\u0645\u062c \u0633\u0637\u062d \u0627\u0644\u0645\u0643\u062a\u0628 \u0648\u0623\u062f\u0627\u0629 Ollama \u062a\u062a\u0637\u0644\u0628 \u062a\u0646\u0633\u064a\u0642 GGUF\u060c \u0628\u064a\u0646\u0645\u0627 \u062a\u062a\u0637\u0644\u0628 \u0623\u0646\u0638\u0645\u0629 \u0627\u0644\u062e\u062f\u0645\u0629 \u0627\u0644\u0642\u0627\u0626\u0645\u0629 \u0639\u0644\u0649 \u0627\u0644\u0640 GPU \u0648\u0623\u064a \u0639\u0645\u0644\u064a\u0629 \u062a\u062a\u0636\u0645\u064e\u0651\u0646 \u0636\u0628\u0637 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u062f\u0642\u064a\u0642 (fine-tuning) \u062a\u0646\u0633\u064a\u0642 safetensors. \u0648\u0639\u0627\u062f\u0629\u064b \u0645\u0627 \u064a\u0646\u0634\u0631 \u0645\u0642\u062f\u0650\u0651\u0645\u0648 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u062a\u0646\u0633\u064a\u0642 safetensors \u0623\u0648\u0644\u064b\u0627\u060c \u062b\u0645 \u064a\u0642\u0648\u0645 \u0627\u0644\u0645\u062c\u062a\u0645\u0639 \u0628\u062a\u062d\u0648\u064a\u0644\u0647 \u0625\u0644\u0649 \u062a\u0646\u0633\u064a\u0642 GGUF \u062e\u0644\u0627\u0644 \u0623\u064a\u0627\u0645 \u0642\u0644\u064a\u0644\u0629.<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Where_GGUF_files_come_from\"><\/span>\u0645\u0635\u062f\u0631 \u0645\u0644\u0641\u0627\u062a GGUF<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<p>\u062a\u0642\u0631\u064a\u0628\u064b\u0627 \u062c\u0645\u064a\u0639 \u0645\u0644\u0641\u0627\u062a GGUF \u0645\u0648\u062c\u0648\u062f\u0629 \u0639\u0644\u0649 \u0645\u0646\u0635\u0629 Hugging Face. \u0648\u062a\u0642\u0648\u0645 \u0634\u0631\u0643\u0629 Labs \u0623\u062d\u064a\u0627\u0646\u064b\u0627 \u0628\u0646\u0634\u0631 \u0645\u0644\u0641\u0627\u062a GGUF \u0631\u0633\u0645\u064a\u0629\u060c \u0644\u0643\u0646 \u0645\u0639\u0638\u0645\u0647\u0627 \u064a\u0623\u062a\u064a \u0645\u0646 \u0645\u064f\u0643\u0645\u0650\u0651\u0645\u064a \u0627\u0644\u0645\u062c\u062a\u0645\u0639 \u2014 \u0645\u062b\u0644 \u0627\u0644\u062d\u0633\u0627\u0628\u0627\u062a <code>bartowski<\/code>, <code>unsloth<\/code>, <code>lmstudio-community<\/code> \u0648 <code>ggml-org<\/code> \u2014 \u0627\u0644\u0630\u064a\u0646 \u064a\u0642\u0648\u0645\u0648\u0646 \u0628\u062a\u062d\u0648\u064a\u0644 \u0643\u0644 \u0625\u0635\u062f\u0627\u0631 \u062c\u062f\u064a\u062f \u0625\u0644\u0649 \u0643\u0627\u0645\u0644 \u0646\u0637\u0627\u0642 \u0645\u0633\u062a\u0648\u064a\u0627\u062a \u0627\u0644\u062a\u0643\u0645\u064a\u0645. \u0648\u0633\u062a\u062d\u062a\u0648\u064a \u0645\u0633\u062a\u0648\u062f\u0639 (repo) \u0628\u0627\u0633\u0645\u064d \u0643\u0640 <code>Meta-Llama-3.1-8B-Instruct-GGUF<\/code> \u0639\u0644\u0649 \u0645\u0644\u0641 \u0648\u0627\u062d\u062f \u0644\u0643\u0644 \u0645\u0633\u062a\u0648\u0649 \u062a\u0643\u0645\u064a\u0645.<\/p>\n<p>\u0648\u0628\u0625\u0645\u0643\u0627\u0646\u0643 \u0623\u064a\u0636\u064b\u0627 \u062a\u062d\u0648\u064a\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0628\u0646\u0641\u0633\u0643 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0623\u062f\u0648\u0627\u062a llama.cpp: <code>convert_hf_to_gguf.py<\/code> \u064a\u062d\u0648\u0650\u0651\u0644 \u0646\u0645\u0648\u0630\u062c Hugging Face \u0625\u0644\u0649 \u0645\u0644\u0641 GGUF \u0628\u0635\u064a\u063a\u0629 F16\u060c \u0623\u0645\u0627 \u0623\u062f\u0627\u0629 <code>llama-quantize<\/code> \u0627\u0644\u062b\u0646\u0627\u0626\u064a\u0629 (binary) (\u0627\u0644\u062a\u064a \u062a\u064f\u0628\u0646\u0649 \u0639\u0646\u062f \u062a\u062c\u0645\u064a\u0639 llama.cpp) \u0641\u062a\u0636\u063a\u0637\u0647 \u0625\u0644\u0649 \u0645\u0633\u062a\u0648\u0649 \u0627\u0644\u062a\u0643\u0645\u064a\u0645 \u0627\u0644\u0630\u064a \u062a\u062e\u062a\u0627\u0631\u0647\u060c \u0645\u062b\u0627\u0644: <code>\u0646\u0645\u0648\u0630\u062c llama-quantize model-f16.gguf model-Q4_K_M.gguf \u0628\u0643\u0645\u064a\u0629 \u0643\u0645\u064a\u0629 Q4_K_M<\/code>.<\/p>\n<h3>\u062a\u062d\u0645\u064a\u0644 \u0645\u0644\u0641 GGUF \u0641\u064a Ollama<\/h3>\n<p>\u064a\u0645\u0643\u0646 \u0644\u0628\u0631\u0646\u0627\u0645\u062c Ollama \u0633\u062d\u0628 \u0645\u0633\u062a\u0648\u062f\u0639\u0627\u062a GGUF \u0645\u0628\u0627\u0634\u0631\u0629\u064b \u0645\u0646 \u0645\u0646\u0635\u0629 Hugging Face\u060c \u0645\u0639 \u062a\u062d\u062f\u064a\u062f \u0646\u0648\u0639 \u0627\u0644\u0643\u0645\u0651\u064a\u0629 \u0628\u0639\u062f \u0627\u0644\u0631\u0645\u0632 \u0627\u0644\u0646\u0642\u0637\u064a (:)<\/p>\n<pre><code>ollama run hf.co\/bartowski\/Meta-Llama-3.1-8B-Instruct-GGUF:Q4_K_M<\/code><\/pre>\n<p>\u0644\u0645\u0644\u0641\u064d \u0644\u062f\u064a\u0643 \u0628\u0627\u0644\u0641\u0639\u0644 \u0639\u0644\u0649 \u0627\u0644\u0642\u0631\u0635 \u0627\u0644\u0635\u0644\u0628\u060c \u0623\u0646\u0634\u0626 \u0645\u0644\u0641 \u0646\u0635 \u0639\u0627\u062f\u064a \u0628\u0627\u0633\u0645 <code>Modelfile<\/code> \u0648\u064a\u062d\u062a\u0648\u064a \u0639\u0644\u0649 \u0645\u0627 \u064a\u0644\u064a:<\/p>\n<pre><code>FROM .\/my-model.gguf<\/code><\/pre>\n<p>\u062b\u0645 \u0633\u062c\u0651\u0644\u0647 \u0648\u0634\u063a\u0651\u0644\u0647 \u0639\u0628\u0631 \u0627\u0644\u0623\u0648\u0627\u0645\u0631 \u0627\u0644\u062a\u0627\u0644\u064a\u0629:<\/p>\n<pre><code>ollama create my-model -f Modelfile\nollama run my-model<\/code><\/pre>\n<p>\u0627\u0644\u0625\u0635\u062f\u0627\u0631\u0627\u062a \u0627\u0644\u062d\u0627\u0644\u064a\u0629 \u0645\u0646 Ollama \u062a\u0642\u0631\u0623 \u0642\u0627\u0644\u0628 \u0627\u0644\u0645\u062d\u0627\u062f\u062b\u0629 \u0627\u0644\u0645\u0636\u0645\u064e\u0651\u0646 \u0641\u064a \u0628\u064a\u0627\u0646\u0627\u062a \u062a\u0639\u0631\u064a\u0641 \u0645\u0644\u0641 GGUF\u061b \u0648\u0625\u0630\u0627 \u0623\u0646\u062a\u062c \u0646\u0645\u0648\u0630\u062c \u0645\u0633\u062a\u0648\u0631\u062f \u0645\u062e\u0631\u062c\u0627\u062a \u0645\u0634\u0648\u064e\u0651\u0634\u0629\u060c \u0641\u0625\u0646 \u0627\u0644\u062d\u0644 \u0639\u0627\u062f\u0629\u064b \u0647\u0648 \u0625\u0636\u0627\u0641\u0629 \u0633\u0637\u0631 \u0635\u0631\u064a\u062d \u064a\u062d\u0645\u0644 \u0627\u0633\u0645 <code>TEMPLATE<\/code> \u0625\u0644\u0649 \u0645\u0644\u0641 Modelfile. \u0623\u0645\u0627 \u062f\u0644\u064a\u0644 Ollama \u0627\u0644\u0634\u0627\u0645\u0644 <a href=\"https:\/\/convly.ai\/ar\/what-is-ollama-complete-guide-2026\/\">\u0641\u064a\u062a\u0646\u0627\u0648\u0644 \u0645\u0644\u0641\u0627\u062a Modelfile \u0628\u0634\u0643\u0644 \u0645\u0639\u0645\u064e\u0651\u0642.<\/a> \u062a\u062d\u0645\u064a\u0644 \u0645\u0644\u0641 GGUF \u0641\u064a LM Studio<\/p>\n<h3>\u0627\u0644\u0637\u0631\u064a\u0642\u0629 \u0627\u0644\u0627\u0639\u062a\u064a\u0627\u062f\u064a\u0629 \u0647\u064a \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0628\u0631\u0646\u0627\u0645\u062c \u0627\u0644\u062a\u0646\u0632\u064a\u0644 \u0627\u0644\u0645\u062f\u0645\u062c: \u0627\u0641\u062a\u062d \u0639\u0644\u0627\u0645\u0629 \u0627\u0644\u062a\u0628\u0648\u064a\u0628 \u00ab\u0627\u0643\u062a\u0634\u0641\u00bb (Discover)\u060c \u0648\u0627\u0628\u062d\u062b \u0639\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0645\u0637\u0644\u0648\u0628\u060c \u0641\u062a\u0639\u0631\u0636 \u0644\u0643 LM Studio \u0642\u0627\u0626\u0645\u0629\u064b \u0628\u0643\u0645\u064a\u0627\u062a GGUF \u0627\u0644\u0645\u062a\u0627\u062d\u0629\u060c \u0645\u0639 \u0627\u0644\u0625\u0634\u0627\u0631\u0629 \u0625\u0644\u0649 \u062a\u0644\u0643 \u0627\u0644\u062a\u064a \u062a\u062a\u0648\u0627\u0641\u0642 \u0645\u0639 \u0623\u062c\u0647\u0632\u062a\u0643. \u0648\u0644\u0627\u0633\u062a\u064a\u0631\u0627\u062f \u0645\u0644\u0641\u064d \u0644\u062f\u064a\u0643 \u0628\u0627\u0644\u0641\u0639\u0644\u060c \u0627\u0633\u062a\u062e\u062f\u0645 \u0648\u0627\u062c\u0647\u0629 \u0633\u0637\u0631 \u0627\u0644\u0623\u0648\u0627\u0645\u0631 (CLI) \u0639\u0628\u0631 \u0627\u0644\u0623\u0645\u0631<\/h3>\n<p>lms import path\/to\/model.gguf<code>\u0623\u0648 \u0627\u0646\u0642\u0644 \u0627\u0644\u0645\u0644\u0641 \u0625\u0644\u0649 \u0645\u062c\u0644\u062f \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u062e\u0627\u0635 \u0628\u0640 LM Studio \u2014 \u062d\u064a\u062b \u064a\u062e\u062a\u0644\u0641 \u0627\u0644\u0645\u0633\u0627\u0631 \u0627\u0644\u0627\u0641\u062a\u0631\u0627\u0636\u064a \u0627\u0644\u062f\u0642\u064a\u0642 \u062d\u0633\u0628 \u0627\u0644\u0625\u0635\u062f\u0627\u0631 \u0648\u0646\u0638\u0627\u0645 \u0627\u0644\u062a\u0634\u063a\u064a\u0644\u060c \u0644\u0643\u0646 \u0639\u0644\u0627\u0645\u0629 \u0627\u0644\u062a\u0628\u0648\u064a\u0628 \u00ab\u0646\u0645\u0627\u0630\u062c\u064a\u00bb (My Models) \u062a\u0639\u0631\u0636 \u0647\u0630\u0627 \u0627\u0644\u0645\u0633\u0627\u0631 \u0648\u062a\u0633\u0645\u062d \u0644\u0643 \u0628\u062a\u063a\u064a\u064a\u0631\u0647. \u0648\u064a\u062c\u0628 \u0623\u0646 \u062a\u0643\u0648\u0646 \u0627\u0644\u0645\u0644\u0641\u0627\u062a \u0645\u0648\u062c\u0648\u062f\u0629 \u0636\u0645\u0646 \u0647\u064a\u0643\u0644 \u0645\u062c\u0644\u062f\u0627\u062a \u0641\u0631\u0639\u064a\u0629 \u0639\u0644\u0649 \u0627\u0644\u0634\u0643\u0644 \u0627\u0644\u062a\u0627\u0644\u064a:<\/code>publisher\/model-name\/ <code>\u0644\u0643\u064a \u064a\u062a\u0645 \u0627\u0644\u062a\u0639\u0631\u064f\u0651\u0641 \u0639\u0644\u064a\u0647\u0627. \u0631\u0627\u062c\u0639<\/code> \u0627\u0644\u062f\u0644\u064a\u0644 \u0627\u0644\u0634\u0627\u0645\u0644 \u0644\u0640 LM Studio <a href=\"https:\/\/convly.ai\/ar\/lm-studio-complete-guide-2026\/\">\u0644\u0644\u0627\u0637\u0644\u0627\u0639 \u0639\u0644\u0649 \u0627\u0644\u0634\u0631\u062d \u0627\u0644\u0643\u0627\u0645\u0644 \u062e\u0637\u0648\u0629\u064b \u0628\u062e\u0637\u0648\u0629.<\/a> \u0647\u0644 \u062a\u0646\u0633\u064a\u0642 GGUF \u0645\u0642\u0635\u0648\u0631\u064c \u0639\u0644\u0649 \u0648\u062d\u062f\u0629 \u0627\u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0645\u0631\u0643\u0632\u064a\u0629 (CPU) \u0641\u0642\u0637\u060c \u0623\u0645 \u0623\u0646\u0647 \u064a\u0633\u062a\u062e\u062f\u0645 \u0623\u064a\u0636\u064b\u0627 \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a (GPU)\u061f<\/p>\n<h2><span class=\"ez-toc-section\" id=\"Frequently_asked_questions\"><\/span>\u0627\u0644\u0623\u0633\u0626\u0644\u0629 \u0627\u0644\u0634\u0627\u0626\u0639\u0629<span class=\"ez-toc-section-end\"><\/span><\/h2>\n<h3>\u0643\u0644\u0627\u0647\u0645\u0627. \u0641\u0642\u062f \u0648\u064f\u0644\u062f \u0645\u0634\u0631\u0648\u0639 llama.cpp \u0641\u064a \u0627\u0644\u0623\u0635\u0644 \u0643\u0645\u0634\u0631\u0648\u0639 \u0627\u0633\u062a\u0646\u062a\u0627\u062c \u064a\u0639\u0645\u0644 \u0639\u0644\u0649 \u0648\u062d\u062f\u0629 \u0627\u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0645\u0631\u0643\u0632\u064a\u0629 \u0641\u0642\u0637\u060c \u0644\u0643\u0646\u0647 \u064a\u064f\u0641\u0648\u0650\u0651\u0636 \u0637\u0628\u0642\u0627\u062a \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0625\u0644\u0649 \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a (\u0639\u0628\u0631 CUDA \u0623\u0648 Metal \u0623\u0648 Vulkan \u0623\u0648 ROCm)\u060c \u0648\u064a\u0639\u0645\u0644 \u0628\u0627\u0644\u0643\u0627\u0645\u0644 \u0639\u0644\u0649 \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a \u0639\u0646\u062f\u0645\u0627 \u064a\u062a\u0633\u0639 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0643\u0627\u0645\u0644\u0627\u064b \u0641\u064a \u0630\u0627\u0643\u0631\u0629 VRAM. \u0648\u062a\u062a\u0639\u0627\u0645\u0644 \u0628\u0631\u0627\u0645\u062c Ollama \u0648LM Studio \u062a\u0644\u0642\u0627\u0626\u064a\u064b\u0651\u0627 \u0645\u0639 \u062a\u0642\u0633\u064a\u0645 \u0627\u0644\u062d\u0645\u0644 \u0628\u064a\u0646 \u0648\u062d\u062f\u0629 \u0627\u0644\u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0645\u0631\u0643\u0632\u064a\u0629 \u0648\u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a\u061b \u0623\u0645\u0627 \u0639\u0646\u062f \u0627\u0633\u062a\u062e\u062f\u0627\u0645 llama.cpp \u0645\u0628\u0627\u0634\u0631\u0629\u064b\u060c \u0641\u064a\u0645\u0643\u0646\u0643 \u0627\u0644\u062a\u062d\u0643\u0645 \u0628\u0647\u0630\u0627 \u0627\u0644\u062a\u0642\u0633\u064a\u0645 \u0628\u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u0627\u0644\u0639\u0644\u0645<\/h3>\n<p>-ngl <code>(\u0639\u062f\u062f \u0637\u0628\u0642\u0627\u062a \u0648\u062d\u062f\u0629 \u0645\u0639\u0627\u0644\u062c\u0629 \u0627\u0644\u0631\u0633\u0648\u0645\u0627\u062a).<\/code> \u0645\u0627 \u0627\u0644\u0641\u0631\u0642 \u0628\u064a\u0646 Q4_K_M \u0648Q4_K_S\u061f<\/p>\n<h3>\u0643\u0644\u0627 \u0627\u0644\u0646\u0648\u0639\u064a\u0646 \u064a\u0645\u062b\u0644\u0627\u0646 \u0643\u0645\u0651\u064a\u0627\u062a k-quants \u0628\u0639\u0631\u0636 4 \u0628\u062a \u062a\u0642\u0631\u064a\u0628\u064b\u0627\u060c \u0648\u0627\u0644\u062d\u0631\u0641 \u064a\u0634\u064a\u0631 \u0625\u0644\u0649 \u0645\u062a\u063a\u064a\u0631 \u0627\u0644\u062d\u062c\u0645.<\/h3>\n<p>(\u0645\u062a\u0648\u0633\u0637) \u062a\u062d\u062a\u0641\u0638 \u0628\u0645\u0632\u064a\u062f \u0645\u0646 \u0627\u0644\u0645\u062a\u062c\u0647\u0627\u062a \u0627\u0644\u062d\u0633\u0627\u0633\u0629 \u0644\u0644\u062c\u0648\u062f\u0629 \u0628\u062f\u0642\u0629 \u0623\u0639\u0644\u0649 \u0645\u0642\u0627\u0631\u0646\u0629\u064b \u0628\u0646\u0648\u0639 <code>_M<\/code> (\u0635\u063a\u064a\u0631)\u060c \u0648\u0628\u0627\u0644\u062a\u0627\u0644\u064a \u0641\u0647\u064a \u0623\u0643\u0628\u0631 \u0642\u0644\u064a\u0644\u064b\u0627 \u0648\u0623\u0641\u0636\u0644 \u0623\u062f\u0627\u0621\u064b \u0642\u0644\u064a\u0644\u064b\u0627. \u0648\u0627\u0644\u0641\u0631\u0642 \u0636\u0626\u064a\u0644 \u062c\u062f\u064b\u0651\u0627 \u2014 \u0644\u0630\u0627 \u0627\u062e\u062a\u0631 <code>_S<\/code> \u0645\u0627 \u0644\u0645 \u062a\u0643\u0646 \u0628\u0636\u0639\u0629 \u0645\u0626\u0627\u062a \u0645\u064a\u062c\u0627\u0628\u0627\u064a\u062a \u0625\u0636\u0627\u0641\u064a\u0629 \u0647\u064a \u0627\u0644\u0639\u0627\u0645\u0644 \u0627\u0644\u062d\u0627\u0633\u0645 \u0641\u064a \u0625\u0645\u0643\u0627\u0646\u064a\u0629 \u062a\u062d\u0645\u064a\u0644 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0639\u0644\u0649 \u062c\u0647\u0627\u0632\u0643. <code>_M<\/code> \u0645\u0627 \u0645\u062f\u0649 \u0627\u0644\u062e\u0633\u0627\u0631\u0629 \u0627\u0644\u0641\u0639\u0644\u064a\u0629 \u0641\u064a \u0627\u0644\u062c\u0648\u062f\u0629 \u0627\u0644\u0646\u0627\u062a\u062c\u0629 \u0639\u0646 \u0627\u0644\u0643\u0645\u0651\u064a\u0629\u061f<\/p>\n<h3>\u0639\u0646\u062f \u0627\u0644\u0643\u0645\u0651\u064a\u0629 Q8_0 \u0648Q6_K\u060c \u062a\u0643\u0648\u0646 \u0627\u0644\u062e\u0633\u0627\u0631\u0629 \u0641\u0639\u0644\u064a\u064b\u0651\u0627 \u0645\u0639\u062f\u0648\u0645\u0629 \u2014 \u0625\u0630 \u064a\u0635\u0639\u0628 \u062c\u062f\u064b\u0651\u0627 \u0639\u0644\u0649 \u0627\u0644\u0627\u062e\u062a\u0628\u0627\u0631\u0627\u062a \u0627\u0644\u0639\u0645\u064a\u0627\u0621 \u0627\u0644\u062a\u0645\u064a\u064a\u0632 \u0628\u064a\u0646\u0647\u0627 \u0648\u0628\u064a\u0646 \u0627\u0644\u0646\u0645\u0648\u0630\u062c \u0627\u0644\u0623\u0635\u0644\u064a \u0628\u062a\u0646\u0633\u064a\u0642 F16. \u0623\u0645\u0627 \u0639\u0646\u062f \u0627\u0644\u0643\u0645\u0651\u064a\u0629 Q4_K_M \u0641\u062a\u0638\u0647\u0631 \u062e\u0633\u0627\u0631\u0629 \u0635\u063a\u064a\u0631\u0629 \u064a\u0645\u0643\u0646 \u0642\u064a\u0627\u0633\u0647\u0627\u060c \u0644\u0643\u0646 \u0645\u0639\u0638\u0645 \u0627\u0644\u0645\u0633\u062a\u062e\u062f\u0645\u064a\u0646 \u0644\u0627 \u064a\u0644\u0627\u062d\u0638\u0648\u0646\u0647\u0627 \u0623\u062b\u0646\u0627\u0621 \u0627\u0644\u0645\u062d\u0627\u062f\u062b\u0629\u060c \u0631\u063a\u0645 \u0623\u0646\u0647\u0627 \u0642\u062f \u062a\u0624\u062b\u0631 \u0641\u064a \u0627\u0644\u0645\u0647\u0627\u0645 \u0627\u0644\u062f\u0642\u064a\u0642\u0629 \u0645\u062b\u0644 \u062a\u0648\u0644\u064a\u062f \u0627\u0644\u0623\u0643\u0648\u0627\u062f \u0623\u0648 \u0627\u0644\u0639\u0645\u0644\u064a\u0627\u062a \u0627\u0644\u0631\u064a\u0627\u0636\u064a\u0629. \u0648\u062a\u0635\u0628\u062d \u0627\u0644\u062a\u062f\u0647\u0648\u0631\u0627\u062a \u0648\u0627\u0636\u062d\u0629 \u0639\u0646\u062f \u0627\u0644\u0643\u0645\u0651\u064a\u0627\u062a \u062f\u0648\u0646 Q4\u060c \u0643\u0645\u0627 \u0623\u0646 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0635\u063a\u064a\u0631\u0629 \u062a\u062a\u0623\u062b\u0631 \u0623\u0643\u062b\u0631 \u0645\u0646 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0643\u0628\u064a\u0631\u0629.<\/h3>\n<p>\u0647\u0644 \u064a\u0645\u0643\u0646\u0646\u064a \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u062a\u0646\u0633\u064a\u0642 GGUF \u0645\u0639 \u0645\u0643\u062a\u0628\u0627\u062a transformers \u0623\u0648 vLLM\u061f<\/p>\n<h3>\u062a\u0648\u062c\u062f \u062f\u0639\u0645 \u0645\u062d\u062f\u0648\u062f \u0644\u0647\u0630\u0627 \u0627\u0644\u0627\u0633\u062a\u062e\u062f\u0627\u0645\u060c \u0644\u0643\u0646\u0647 \u0644\u064a\u0633 \u0627\u0644\u0637\u0631\u064a\u0642 \u0627\u0644\u0642\u064a\u0627\u0633\u064a \u0623\u0648 \u0627\u0644\u0623\u0635\u0644\u064a \u2014 \u0641\u0645\u0643\u062a\u0628\u0629 transformers \u062a\u0633\u062a\u0637\u064a\u0639 \u0625\u0644\u063a\u0627\u0621 \u0627\u0644\u0643\u0645\u0651\u064a\u0629 (dequantise) \u0644\u0628\u0639\u0636 \u0645\u0644\u0641\u0627\u062a GGUF \u0639\u0646\u062f \u0627\u0644\u062a\u062d\u0645\u064a\u0644\u060c \u0628\u064a\u0646\u0645\u0627 \u062a\u0645\u062a\u0644\u0643 vLLM \u062f\u0639\u0645\u064b\u0627 \u062a\u062c\u0631\u064a\u0628\u064a\u064b\u0651\u0627 \u0644\u0640 GGUF \u064a\u062a\u0641\u0627\u0648\u062a \u062d\u0633\u0628 \u0628\u0646\u064a\u0629 \u0627\u0644\u0646\u0645\u0648\u0630\u062c. \u0641\u0625\u0630\u0627 \u0643\u0646\u062a \u062a\u0628\u0646\u064a \u062a\u0637\u0628\u064a\u0642\u0627\u062a\u0643 \u0639\u0644\u0649 \u0647\u0630\u0647 \u0627\u0644\u0645\u0643\u062a\u0628\u0627\u062a\u060c \u0641\u0645\u0646 \u0627\u0644\u0623\u0641\u0636\u0644 \u0627\u0633\u062a\u062e\u062f\u0627\u0645 \u062a\u0646\u0633\u064a\u0642 safetensors\u061b \u0623\u0645\u0627 GGUF \u0641\u0647\u0648 \u0627\u0644\u062a\u0646\u0633\u064a\u0642 \u0627\u0644\u0623\u0645\u062b\u0644 \u0644\u0623\u0646\u0638\u0645\u0629 \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0627\u0644\u062a\u0627\u0628\u0639\u0629 \u0644\u0639\u0627\u0626\u0644\u0629 llama.cpp.<\/h3>\n<p>\u0644\u0645\u0627\u0630\u0627 \u0627\u0646\u0642\u0633\u0645 \u0646\u0645\u0648\u0630\u062c\u064a \u0625\u0644\u0649 \u0639\u062f\u0629 \u0645\u0644\u0641\u0627\u062a \u0628\u0627\u0645\u062a\u062f\u0627\u062f .gguf\u061f<\/p>\n<h3>\u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0643\u0628\u064a\u0631\u0629 \u062c\u062f\u064b\u0651\u0627 \u062a\u064f\u0642\u0633\u064e\u0651\u0645 \u0625\u0644\u0649 \u0623\u062c\u0632\u0627\u0621 \u062a\u062d\u0645\u0644 \u0623\u0633\u0645\u0627\u0621\u064b \u0645\u062b\u0644<\/h3>\n<p>model-00001-of-00002.gguf <code>\u0648\u063a\u0627\u0644\u0628\u064b\u0627 \u0645\u0627 \u064a\u0643\u0648\u0646 \u0630\u0644\u0643 \u0644\u0636\u0645\u0627\u0646 \u0639\u062f\u0645 \u062a\u062c\u0627\u0648\u0632 \u0627\u0644\u062d\u062f \u0627\u0644\u0623\u0642\u0635\u0649 \u0644\u062d\u062c\u0645 \u0627\u0644\u0645\u0644\u0641 \u0627\u0644\u0645\u0633\u0645\u0648\u062d \u0628\u0647 \u0639\u0644\u0649 \u0645\u0646\u0635\u0627\u062a \u0627\u0644\u0627\u0633\u062a\u0636\u0627\u0641\u0629. \u0648\u064a\u062c\u0628 \u0627\u0644\u0627\u062d\u062a\u0641\u0627\u0638 \u0628\u062c\u0645\u064a\u0639 \u0627\u0644\u0623\u062c\u0632\u0627\u0621 \u0641\u064a \u0627\u0644\u0645\u062c\u0644\u062f \u0646\u0641\u0633\u0647\u060c \u062b\u0645 \u062a\u0648\u062c\u064a\u0647 \u0628\u0631\u0646\u0627\u0645\u062c \u0627\u0644\u062a\u0634\u063a\u064a\u0644 \u0625\u0644\u0649 \u0627\u0644\u0645\u0644\u0641 \u0627\u0644\u0623\u0648\u0644 \u2014 \u0641\u0633\u062a\u0642\u0648\u0645 \u0645\u0643\u062a\u0628\u0629 llama.cpp \u0648\u0627\u0644\u0623\u062f\u0648\u0627\u062a \u0627\u0644\u0645\u0628\u0646\u064a\u0629 \u0639\u0644\u064a\u0647\u0627 \u062a\u0644\u0642\u0627\u0626\u064a\u064b\u0651\u0627 \u0628\u062a\u062d\u0645\u064a\u0644 \u0628\u0627\u0642\u064a \u0627\u0644\u0623\u062c\u0632\u0627\u0621.<\/code>\u0648\u0628\u0645\u062c\u0631\u062f \u0623\u0646 \u062a\u0639\u0631\u0641 \u0623\u064a \u0643\u0645\u0651\u064a\u0629 \u062a\u062a\u0648\u0627\u0641\u0642 \u0645\u0639 \u0623\u062c\u0647\u0632\u062a\u0643\u060c \u0641\u0625\u0646 \u0627\u0644\u0633\u0624\u0627\u0644 \u0627\u0644\u0639\u0645\u0644\u064a \u0627\u0644\u062a\u0627\u0644\u064a \u0647\u0648: \u0623\u064a \u0646\u0645\u0648\u0630\u062c \u064a\u062c\u0628 \u062a\u062d\u0645\u064a\u0644\u0647 \u0641\u064a \u0647\u0630\u0647 \u0627\u0644\u0643\u0645\u0651\u064a\u0629\u061f \u2014 \u0627\u0644\u0642\u0627\u0626\u0645\u0629 \u0627\u0644\u0645\u0639\u0646\u0648\u0646\u0629<\/p>\n<p>\u062a\u0633\u0631\u062f \u0645\u0648\u0627\u0635\u0641\u0627\u062a 37 \u0646\u0645\u0648\u0630\u062c\u064b\u0627 \u062d\u0627\u0644\u064a\u064b\u0651\u0627\u060c \u0648\u0627\u062d\u062a\u064a\u0627\u062c\u0627\u062a\u0647\u0627 \u0645\u0646 \u0630\u0627\u0643\u0631\u0629 VRAM \u0648\u0623\u0633\u0639\u0627\u0631\u0647\u0627\u060c \u0628\u064a\u0646\u0645\u0627 \u062a\u0648\u0641\u0631 \u0642\u0627\u0626\u0645\u0629 <a href=\"https:\/\/convly.ai\/ar\/models\/\">\u0642\u0627\u0639\u062f\u0629 \u0628\u064a\u0627\u0646\u0627\u062a \u0627\u0644\u0646\u0645\u0627\u0630\u062c<\/a> \u0623\u0641\u0636\u0644 \u0627\u0644\u0646\u0645\u0627\u0630\u062c \u0627\u0644\u0645\u062d\u0644\u064a\u0629 \u0644\u0640 Ollama <a href=\"https:\/\/convly.ai\/ar\/best-local-llms-to-run-on-ollama-2026\/\">\u0642\u0627\u0626\u0645\u0629 \u0645\u0648\u062c\u0632\u0629 \u0645\u0645\u062a\u0627\u0632\u0629 \u0644\u062a\u0628\u062f\u0623 \u0645\u0646\u0647\u0627.<\/a> \u0642\u0627\u0626\u0645\u0629 \u062c\u064a\u062f\u0629 \u0642\u0635\u064a\u0631\u0629 \u064a\u0645\u0643\u0646 \u0627\u0644\u0628\u062f\u0621 \u0645\u0646\u0647\u0627.<\/p>","protected":false},"excerpt":{"rendered":"<p>GGUF is the file format for running LLMs locally. It packs a model&#8217;s weights, tokenizer and configuration into a single [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":2104,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"site-sidebar-layout":"default","site-content-layout":"","ast-site-content-layout":"default","site-content-style":"default","site-sidebar-style":"default","ast-global-header-display":"","ast-banner-title-visibility":"","ast-main-header-display":"","ast-hfb-above-header-display":"","ast-hfb-below-header-display":"","ast-hfb-mobile-header-display":"","site-post-title":"","ast-breadcrumbs-content":"","ast-featured-img":"","footer-sml-layout":"","ast-disable-related-posts":"","theme-transparent-header-meta":"","adv-header-id-meta":"","stick-header-meta":"","header-above-stick-meta":"","header-main-stick-meta":"","header-below-stick-meta":"","astra-migrate-meta-layouts":"default","ast-page-background-enabled":"default","ast-page-background-meta":{"desktop":{"background-color":"var(--ast-global-color-5)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"ast-content-background-meta":{"desktop":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"tablet":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""},"mobile":{"background-color":"var(--ast-global-color-4)","background-image":"","background-repeat":"repeat","background-position":"center center","background-size":"auto","background-attachment":"scroll","background-type":"","background-media":"","overlay-type":"","overlay-color":"","overlay-opacity":"","overlay-gradient":""}},"footnotes":""},"categories":[7],"tags":[],"class_list":["post-2103","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-news"],"_links":{"self":[{"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/posts\/2103","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/comments?post=2103"}],"version-history":[{"count":1,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/posts\/2103\/revisions"}],"predecessor-version":[{"id":2105,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/posts\/2103\/revisions\/2105"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/media\/2104"}],"wp:attachment":[{"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/media?parent=2103"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/categories?post=2103"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/convly.ai\/ar\/wp-json\/wp\/v2\/tags?post=2103"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}