Technology · 95 views
Kimi K2.7-Code cuts thinking tokens 30% — but practitioners say the benchmarks don't check out
Moonshot AI released Kimi K2.7-Code this week, an open-source update to its K2 coding model family, claiming leaner reasoning and double-digit performance gains.
AI Summary
Moonshot AI unveiled Kimi K2.7-Code, an open‑source update to its K2 coding model family that uses the same trillion‑parameter mixture‑of‑experts architecture as the earlier K2.6 model. The company claims the new version reduces “thinking‑token” usage by 30%, which should lower inference costs for teams running agentic workflows. K2.7-Code is released under a Modified MIT license, with weights hosted on HuggingFace and can be deployed via vLLM or SGLang. Practitioners have begun questioning whether the reported efficiency gains hold up on independent benchmarks.
AI summaries can be wrong sometimes—always verify important details using the source article.
How AI & Automation are usedMore from Technology
Continue reading recent Technology coverage
- Anthropic is letting METR researchers look under the hood with employee-level accessContinue reading
- Bluesky expands its AI-powered Attie app to let anyone build websites, games, and moreContinue reading
- Meta and Apple team up to make Muse AI models run better on MacsContinue reading
- Last chance to save up to $200 on your TechCrunch Founder Summit pass: Early booking ends October 16Continue reading
Support HappeningNow
Independent AI-powered news analysis is reader-supported. Your contribution helps cover infrastructure, summaries, and continued platform development.
Support HappeningNow