KwaiKAT releases KAT-Coder-Pro V2: non-reasoning code model matches Claude Sonnet 4.6
Original title:KwaiKAT has released KAT-Coder-Pro V2, a non-reasoning model that scores 44 on the Artificial Analys...
AI summary
KwaiKAT released the non-reasoning code model KAT-Coder-Pro V2, which scored 44 on the Artificial Analysis Intelligence Index, up 8 points from V1 and on par with Claude Sonnet 4.6. The model is notably token-efficient, running on only about 9M output tokens, far below the Claude series and reasoning models such as DeepSeek. Agent capabilities improved substantially, with Terminal-Bench Hard at 49% (up 40 percentage points), matching Claude Opus 4.6. Cost fell to $73 and response speed reached 109 tokens/sec. However, it regressed from V1 on long-context reasoning and knowledge recall.
Key Takeaways
- KwaiKAT released the non-reasoning code model KAT-Coder-Pro V2,which scored 44 on the Artificial…
- The model is notably token-efficient,running on only about 9M output tokens,far below the Claude…
- Agent capabilities improved substantially,with Terminal-Bench Hard at 49% (up 40 percentage…
- Cost fell to $73 and response speed reached 109 tokens/sec.
- However,it regressed from V1 on long-context reasoning and knowledge recall.
Quick Facts
- Published
- Updated
- Category
- Other
快手发布 KAT-Coder-Pro V2,非推理架构实现 44 分智能指数,Agent 能力跃升 40 个百分点,成本仅为 Claude Sonnet 的 5%。
Related entities
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: KwaiKAT releases KAT-Coder-Pro V2: non-reasoning code model matches Claude Sonnet 4.6. March 31, 2026. https://new.nat6.net/en/news/cmnw1ypby00odslc3gajfm7x6/