DeepSeek-V4-Pro-0813
#24 in Open-Source LLMsDeepSeek · v4 · pro 0813 · since 12./13. August 2026 · 5× · last seen Aug 17, 2026
46
Momentum
DeepSeek-V4-Pro-0813 is the official general-availability release of DeepSeek-V4-Pro, superseding the April 2026 preview version. It is a mixture-of-experts language model with roughly 1.6 trillion total parameters and about 49 billion active parameters per token, supporting a context window of 1,048,576 tokens. Compared to the preview, the model shows greatly enhanced agentic capabilities, particularly in coding, tool-use, and cybersecurity benchmarks, and is released as open weights under the MIT license. It is accessible via the DeepSeek API, DeepSeek Chat, Hugging Face, and third-party providers such as OpenRouter.
Momentum trend
19.05.17.08.
Features
| Key Benchmark (%) | Terminal-Bench 2.1: 87.9% (own DeepSeek harness); SWE-bench Verified: 96.40% (Vals AI, #2 of 82 models) |
| Context Window (Tokens) | 1,048,576 tokens (1M), max output 384,000 tokens |
| License | MIT License (open weights) |
| Multimodality | No image/vision support (text-only model) |
| Platform | DeepSeek API, DeepSeek Chat, Hugging Face, OpenRouter, Fireworks AI |
| Price per 1M Tokens | $0.435 input (cache miss) / $0.003625 input (cache hit) / $0.87 output |
| Release Date | August 12, 2026 (API rollout), model card published August 13, 2026 |