DeepSeek-V4-Pro-Max
#119 in Frontier LLMsDeepSeek · v4 · pro max · 2× · last seen Jul 30, 2026
16
Momentum
DeepSeek-V4-Pro-Max is the maximum Reasoning mode of DeepSeek-V4-Pro, a Mixture-of-Experts language model from DeepSeek with 1.6 trillion total parameters (49 billion activated) and a 1-million-Token context window. Released on April 24, 2026 as an Open-Weight model under the MIT license, it is positioned by DeepSeek as the best available Open-Source model, achieving top scores in Coding and Knowledge Benchmarks, approaching leading Closed-Source models. Access is available via the DeepSeek API, chat.deepseek.com, and various third-party providers (including OpenRouter, DeepInfra, and NVIDIA NIM).
Momentum trend
01.05.30.07.
Features
| Key Benchmark (%) | GPQA Diamond: 90.1% (Pass@1); MMLU-Pro: 87.5% (EM) |
| Context Window (Tokens) | 1,048,576 tokens (1M) |
| License | MIT License (open weights) |
| Multimodality | Text-only (input and output are text only) |
| Platform | DeepSeek API (api.deepseek.com), chat.deepseek.com, Hugging Face open weights, third-party providers (OpenRouter, DeepInfra, NVIDIA NIM, Ollama, etc.) |
| Price per 1M Tokens | $0.435 input / $0.87 output (cache-hit input $0.003625) |
| Release Date | April 24, 2026 |