Qwen3.8-27B — Specifications
Compiled by Mustafa Ihsan from the vendor’s published documentation · Last updated
| Developer | Alibaba |
|---|---|
| Type | LLM (dense, multimodal) |
| Modality | Text, Image, Video → Text |
| Parameters | 27B |
| Context window | 256K native (up to 1M) |
| Max output | 128K |
| License | Apache 2.0 (open) |
| Open weights | Yes |
| Released | 2026-08 |
| Input price | $0.50 /1M |
| Output price | $3.00 /1M |
| API providers | Alibaba Cloud Model Studio (International), OpenRouter; Ollama (local) |
Run it locally
| VRAM (4-bit) | ~18 GB (Ollama 4-bit build) |
|---|
Benchmarks
| Qwen-reported | Terminal Bench 2.1 (Terminus) 73.0; SWE-bench Pro 61.7 (Qwen3.6-27B: 53.5); NL2Repo-Bench 42.3. Prices shown are Alibaba Cloud Model Studio (International). |
|---|
What is Qwen3.8-27B?
Qwen3.8-27B is the dense, 27-billion-parameter member of Alibaba’s Qwen3.8 family and a native vision-language model: it reads text, images and video and writes text. The weights are on Hugging Face under Apache 2.0, and Alibaba added it to Model Studio (International) on 17 August 2026. It has a 262,144-token native context that Qwen says extends to 1,000,000 tokens, and it thinks by default, though thinking can be switched off per request. Alibaba says it improves on Qwen3.6-27B mainly in coding and office tasks; Qwen reports 61.7 on SWE-bench Pro against 53.5 for Qwen3.6-27B.
Alibaba Cloud Model Studio (International) charges $0.50 per million input tokens and $3.00 per million output tokens, with implicit cache hits at $0.10 per million, and serves it with a 1,000,000-token context and up to 131,072 output tokens. Those are the prices on this page. Qwen says a hosted version on its Qwen Cloud service is coming soon.
For local use, Qwen’s own downloads are the BF16 weights, about 55.6 GB, and an FP8 version. Ollama’s library carries 4-bit builds such as qwen3.8:27b-q4_K_M, an 18 GB download, so plan on a 24 GB GPU. It is the modern counterpart to the April 2025 Qwen3 32B and a step up in size from Qwen3 14B, adding image and video input and a much longer context. For Alibaba’s hosted flagship, see Qwen3.8-Max. Check your quantisation and context length with the LLM VRAM calculator.
Qwen3.8-27B pricing: API cost per 1M tokens
| Input (per 1M tokens) | $0.500 |
|---|---|
| Output (per 1M tokens) | $3.00 |
| Output/input ratio | 6× |
| Blended (4:1 in:out) | $1.00 per 1M tokens |
What Qwen3.8-27B costs per month
Real monthly spend at a 4:1 input-to-output mix — the ratio a typical chat or RAG workload actually produces.
| Workload | Tokens / month | Cost / month |
|---|---|---|
| Side project | 1M in / 0.25M out | $1.25 |
| Small team | 20M in / 5M out | $25 |
| Production | 200M in / 50M out | $250 |
Run your own numbers in the AI API cost calculator.
Cheaper alternatives to Qwen3.8-27B
| Model | Blended $/1M | You save |
|---|---|---|
| Llama 3.1 8B open | $0.0240 | 98% cheaper |
| Mistral NeMo 12B open | $0.0292 | 97% cheaper |
| Gemma 3 4B open | $0.0600 | 94% cheaper |
Self-host or pay the API?
Qwen3.8-27B is open-weight, so you can run it yourself. It needs ~18 GB (Ollama 4-bit build) of VRAM at 4-bit. Self-hosting only beats the API once your volume is high enough to keep that hardware busy — the self-hosting vs API calculator works out the break-even point for your token volume.
Frequently asked questions
How much does Qwen3.8-27B cost per 1M tokens?
Qwen3.8-27B costs $0.500 per 1M input tokens and $3.00 per 1M output tokens. At a typical 4:1 input-to-output mix that blends to about $1.00 per 1M tokens.
How much does Qwen3.8-27B cost per month?
A small-team workload of 20M input and 5M output tokens a month costs about $25 on Qwen3.8-27B. A side project (1M in / 0.25M out) costs roughly $1.25.
What is a cheaper alternative to Qwen3.8-27B?
Llama 3.1 8B is the strongest cheaper option in our database at $0.0240 per 1M blended — about 98% less than Qwen3.8-27B. It is also open-weight, so self-hosting is an option.
Can I run Qwen3.8-27B locally?
Yes. Qwen3.8-27B is open-weight and needs about ~18 GB (Ollama 4-bit build) of VRAM at 4-bit quantisation.
Why does Qwen3.8-27B charge more for output than input?
Output tokens are generated one at a time and cannot be batched the way a prompt can, so they cost the provider more to serve. Qwen3.8-27B charges 6× more for output, which is why prompt-heavy workloads are far cheaper to run than generation-heavy ones.
Prices are the published list rates for the model's primary API and are reviewed as providers change them. Volume, batch and cached-input discounts are not included. Compare every model side by side in the AI models database or the LLM leaderboard.