

DeepSeek models
#47 in Open-Source LLMsDeepSeek · since 2026-04-24 · 2× · last seen Nov 19, 2026
19
Momentum
DeepSeek V4 is a family of open-weight Mixture-of-Experts language models from the Chinese company DeepSeek, announced on April 24, 2026 with V4-Pro and V4-Flash. Both support a 1M-token context window, and the weights are published on Hugging Face under the MIT license. API prices are listed on DeepSeek's official pricing page and may change.
Momentum trend
24.08.22.11.
Features
| Key Benchmark (%) | SWE-bench Verified: 80.6% (V4-Pro, vendor-reported in model card) |
| Context Window (Tokens) | 1,000,000 tokens (V4-Pro and V4-Flash) |
| License | MIT (code and weights) |
| Multimodality | V4.1-Flash: vision supported; V4-Pro: no vision (text and code) |
| Platform | API (OpenAI Chat Completions and Anthropic-compatible), weights on Hugging Face |
| Price per 1M Tokens | V4.1-Flash: input (cache miss) $0.15, output $0.60 (off-peak); V4-Pro: input $0.66, output $1.98 (off-peak) |
| Release Date | April 24, 2026 (V4 Preview); V4-Pro-0813 on August 13, 2026; V4.1-Flash GA on September 9, 2026 |