DeepSeek-V4-Pro

weights

open(MIT, on HF, both Pro and Flash)

data

closed

code

partial(inference/serving

license

MIT(permissive, no use restrictions)

MIT weights across the V4 line, training corpus not released, inference and serving code only, so 3. V4-Pro governs and both it and V4-Pro-Base are plain MIT, which keeps the tier clear of an unsettled question: the retired deepseek-v3-base record carried a compound `code MIT + model DeepSeek-Model-License`, and whether that license's acceptable-use restrictions count as use-restricting is open in issue #117. The tier does not depend on that ruling. Adoption 5 is carried from V3.2, the most downloaded release.

• https://huggingface.co/api/models/deepseek-ai/DeepSeek-V4-Pro recorded 2026-07-29

  license mit, ungated, 864.8GB of weight files, 1,635,092 downloads last month

• https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro recorded 2026-06-04

  flagship phase-C verification source

• https://api-docs.deepseek.com/news/news260424 recorded 2026-06-04

  flagship phase-C verification source

• https://simonwillison.net/2026/apr/24/deepseek-v4/ recorded 2026-06-04

  flagship phase-C verification source

ATOM Report (Apr 2026): China open models reached ~1.15B cumulative HF downloads with DeepSeek a primary driver; DeepSeek-V3.2 widely served (OpenRouter, multiple inference providers) with sustained post-launch production usage and frontier-at-10x-lower-cost positioning. Real usage well into the >10M-equivalent band across web app + API + derivatives.

• https://arxiv.org/html/2604.07190v1 recorded 2026-06-04

  flagship phase-C verification source

• https://openrouter.ai/deepseek/deepseek-v3.2 recorded 2026-06-04

  flagship phase-C verification source

• https://introl.com/blog/deepseek-v3-2-open-source-ai-cost-advantage recorded 2026-06-04

  flagship phase-C verification source

SWE-bench Verified 80.6% (trails Claude Opus 4.6 by ~0.2), MMLU-Pro 87.5, GPQA Diamond 90.1, LiveCodeBench Pass@1 93.5 (reported best of any model). V4-Pro-Max described as the strongest open source model available, at/near closed frontier on coding and reasoning.

• https://codersera.com/blog/deepseek-v4-pro-review-benchmarks-pricing-2026/ recorded 2026-06-04

  flagship phase-C verification source

• https://benchlm.ai/models/deepseek-v4-pro-high recorded 2026-06-04

  flagship phase-C verification source

• https://www.morphllm.com/deepseek-v4 recorded 2026-06-04

  flagship phase-C verification source