MiniMax launches M3.1 Flash for coding, agentic tasks
MiniMax said Space Bunny, which processed 200T+ tokens anonymously, was an early M3.1 Flash.
Chinese AI startup Hangzhou DeepSeek has introduced an updated V4.1-Flash model named V4.1-Flash-1017, claiming it completes tasks more swiftly. Launched over five weeks after the initial V4.1-Flash release, the minor update is described on its Hugging Face model card as reducing task execution time and output length, though the exact improvement is unspecified.
In September, DeepSeek priced API rates starting at 15 cents per million uncached input tokens. Following the update, shares of Chinese rivals MiniMax Group Inc. and Alibaba Group Holding Ltd. dropped in Hong Kong. DeepSeek's internal testing revealed that the new model achieved 77% success on the DeepSWE v1.1 software engineering test, up from 74.2% for the previous version.
Only seven out of eleven benchmarks DeepSeek published still favor the new model, with a notable 14-point margin against Anthropic PBC's Claude Opus 5. The update is tailored for agent software, including DeepSeek Harness and two Tencent products - CodeBuddy and WorkBuddy. Both agent builders received a second release from DeepSeek.
The company also opened its search engine to outside developers through a Search API returning up to 50 ranked web results per query, and a Fetch API converting web pages to Markdown, both priced at 75 cents per 1,000 successful requests. DeepSeek's in-house Search API and Fetch API were both free until November 1. The release comes after a five-hour API outage on October 15, which DeepSeek apologized for and promised to prevent in the future by removing single points of dependency in its system.
Written by urgent.news from SiliconANGLE's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
This story
This is one outlet's version. Read the fullest account.