TL;DR: DeepSeek released open-weight V4 Flash 0731 under MIT license, ranking among top open models on the Artificial Analysis Intelligence Index.
Summary: DeepSeek published the weights for its new flash-tier model, DeepSeek V4 Flash 0731, under the MIT license, permitting unrestricted commercial use and modification. The model has 284B total parameters (13B active) in mixed FP4/FP8 precision (~167GB) and lands among the top 3 open-weight models on the Artificial Analysis leaderboard. It is also available via DeepSeek's first-party API.
Why it matters: For AI builders, this offers a commercially unrestricted, frontier-adjacent open-weight model at a relatively small active-parameter size, making it attractive for efficient self-hosted deployments. Try benchmarking it against other open models on the Pareto frontier for Intelligence Index vs. total parameters, and consider it for applications needing strong reasoning without prohibitive inference costs.
Source: x_com