DeepSeek V4-Pro — Specifications
| Developer | DeepSeek |
|---|
| Type | LLM (MoE) |
|---|
| Modality | Text → Text |
|---|
| Parameters | 1.6T total / ~49B active (MoE) |
|---|
| Context window | 1M |
|---|
| Max output | 384K |
|---|
| License | MIT (open) |
|---|
| Open weights | ✅ Yes |
|---|
| Released | 2026-04 |
|---|
| Input price | $0.435 /1M |
|---|
| Output price | $0.87 /1M |
|---|
| API providers | DeepSeek, OpenRouter |
|---|
🖥️ Run it locally
| VRAM (4-bit) | ~800 GB |
|---|
| Minimum GPU | Multi-GPU server (e.g. 8× H100 80GB) |
|---|
Official page →
DeepSeek’s open flagship — a 1.6-trillion-parameter mixture-of-experts (≈49B active per token) with a 1M-token context, at a small fraction of frontier pricing. Open weights under the MIT license; the API speaks both OpenAI and Anthropic formats.
