What happened
DeepSeek released V4 Pro 0813, its latest Pro model, available through OpenRouter via API. The company has not published an official announcement page for the model.
Why it matters
DeepSeek's track record of open-sourcing previous Pro versions (V4-Pro in April and V4-Flash-0731 in July) suggests open weights may follow, though the company has not confirmed plans. The model appears to generate visibly different outputs depending on reasoning level (low, medium, high), a behavior the author has not observed in other models.
What to watch
Availability is currently API-only via OpenRouter; benchmark data has circulated informally through WeChat, Reddit, and Hacker News rather than official channels.
Summaries like this, in your inbox every morning.
DeepSeek's release of V4 Pro 0813 continues the company's pattern of iterative Pro model releases—the body mentions April's V4-Pro and July's V4-Flash-0731 as predecessors. The lack of an official announcement suggests a quiet or informal rollout, with benchmark results circulating through community channels (WeChat, Reddit, Hacker News) rather than corporate communication. This mirrors DeepSeek's broader strategy of making models available first and formalizing announcements later. The model's apparent ability to generate visually distinct outputs across reasoning levels (low, medium, high) marks a notable technical characteristic; the author explicitly notes not having observed this behavior in other models, implying it may be a distinguishing feature worth monitoring.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Circlia started a proof-of-concept for TASTE SIGNAL: visitors to Holland Village Cafe FUKUOKA taste cookies an…
In a webinar hosted by French Mathematical Society president Isabelle Gallagher, Fields Medal winner Hugo Dumi…
NVIDIA released Nemotron 3.5 Super VL, an open-weight model with 120 billion parameters that handles text, ima…
NVIDIA released Nemotron 3.5 Bolt 74B-A9B 24GB, an AI model for local agents that hits over 80 tokens per seco…
NVIDIA published a blog post titled "AI Security Is an Engineering Problem - How to Solve It at Every Layer of…
In the test, Haiku 5.5 on medium setting averaged scores close to Sonnet 5.5 on routine documents at about 1/1…
