AI News Feed
Market watch
DeepSeek releases V4.1-Flash, says smaller model beats flagship V4-ProUS Navy Certifies Rolls-Royce Engines for Unmanned Vessels Built to Run 30 Days Without a CrewOracle shares rise as cloud infrastructure revenue jumps 121%Anthropic says AI lets lone operators run state-level hacking campaignsCase report links daily laptop use to 'toasted skin' marks on boy's abdomeniPhone 18 Pro Max and Galaxy S26 Ultra Match on Size and Battery, Split on Cameras and Weight
DeepSeek releases V4.1-Flash, says smaller model beats flagship V4-ProUS Navy Certifies Rolls-Royce Engines for Unmanned Vessels Built to Run 30 Days Without a CrewOracle shares rise as cloud infrastructure revenue jumps 121%Anthropic says AI lets lone operators run state-level hacking campaignsCase report links daily laptop use to 'toasted skin' marks on boy's abdomeniPhone 18 Pro Max and Galaxy S26 Ultra Match on Size and Battery, Split on Cameras and Weight
Large Language Models

DeepSeek releases V4.1-Flash, says smaller model beats flagship V4-Pro

DeepSeek released V4.1-Flash, a 552B open-weight model it says beats its flagship V4-Pro on cost and speed, and will route V4-Pro API calls to it.

AI News Feed · 2026-09-11 09:52

Starting Sept. 14, requests sent to V4-Pro through DeepSeek's application programming interface will be answered by V4.1-Flash and billed at the smaller model's rates until a V4.1-Pro version launches. V4-Flash and the experimental vision model DeepSeek shipped in August are retired, and calls to either now land on V4.1-Flash.

V4.1-Flash is a mixture-of-experts model with 552 billion parameters, close to double the 284 billion in V4-Flash. A causal encoder-decoder design keeps 8 billion parameters active while the model processes a prompt and 16 billion while it generates output. Image understanding, offered only in the August experimental release, is now built into the model itself.

Much of the engineering went into shrinking the key-value cache. According to DeepSeek's technical report, the model stores those entries in a four-bit floating-point format, giving it a global footprint of 890 bytes per token, about a quarter of what V4-Flash requires. Persistent cache storage on SSDs falls to roughly an eighth of the previous generation's.

DeepSeek's benchmark table compares the model at maximum reasoning effort against Anthropic's Claude Opus 5 and OpenAI's GPT-5.6 Sol. V4.1-Flash scored 90.6 on Terminal-Bench 2.1, ahead of Opus 5 at 89.1 and GPT-5.6 Sol at 88.8. On the DeepSWE v1.1 software engineering test it resolved 74.2% of tasks, against 74% for Opus 5 and 62.7% for V4-Pro. Both U.S. models still lead on the GPQA Diamond science reasoning benchmark.

Off-peak API pricing is 15 cents per million uncached input tokens and 60 cents per million output tokens, with rates doubling during weekday peak windows. Developers still calling V4-Pro pay $3.96 per million output tokens at peak, compared with $1.20 for V4.1-Flash, a reduction of roughly 70% on output.

Weights are available on Hugging Face under the MIT license, and the model is live in DeepSeek's web and mobile apps. DeepSeek said it will work with the open-source community on inference support and explore further deployment options.

The launch came the same day Anthropic named DeepSeek in its latest threat intelligence report as one of seven China-based labs it says ran distillation campaigns against Claude. Anthropic attributed more than 12.1 million exchanges over 14 days in July to DeepSeek. DeepSeek has not publicly responded to that report.

DeepSeek grew out of the Chinese hedge fund High-Flyer. Founder Liang Wenfeng reportedly contributed $3 billion to a funding round of more than $7.4 billion in June that valued the company above $50 billion.