Moonshot AI releases Kimi K2.7-Code, an open-source model for coding and agentic workflows that uses 30% fewer reasoning tokens while delivering double-digit benchmark improvements.

// 1h agoMODEL RELEASE

Moonshot AI releases Kimi K2.7-Code, an open-source model for coding and agentic workflows that uses 30% fewer reasoning tokens while delivering double-digit benchmark improvements.

Moonshot AI has announced Kimi K2.7-Code, a major open-source update to their coding model. Compared to Kimi K2.6, the new model achieves double-digit performance gains, including +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite. Crucially, Kimi K2.7-Code achieves these improvements while consuming 30% fewer reasoning tokens, marking a significant efficiency leap.

// ANALYSIS

Efficiency is the new frontier in LLM development; reducing reasoning tokens by 30% while increasing accuracy makes Kimi K2.7-Code a highly practical choice for agentic workflows where API costs and latency are critical.

* Open-source availability allows developers to run and deploy the model locally, bypassing commercial API dependencies.

* Double-digit gains across multiple benchmarks highlight a rapid pace of improvement for the Kimi series.

* The 30% drop in reasoning token usage signals a trend toward smarter, more efficient reasoning architectures rather than purely scaling up.

// TAGS

agentic-workflowsllmmoonshot-aiopen-source

DISCOVERED

1h ago

2026-06-13

PUBLISHED

2h ago

2026-06-13

RELEVANCE

8/ 10

AUTHOR

mark_k

// KEEP READING

More AI developer news from the feed

EXPLORE FULL FEED

UPDATE34m ago

AI Templates updates layout, jobs section

AI Templates (aitmpl.com) has updated its interface, shifting back to its original layout and restructuring its jobs section to support developers building AI-native workflows. The platform serves as a community-driven repository and CLI hub for Anthropic's Claude Code, now hosting over 840 skills, 420 agents, and 80 Model Context Protocol (MCP) integrations, with a new CLI release supporting job integrations coming soon.

NEWS40m ago

Designer claims access to suspended Claude Fable 5

Designer Alex Socoloff claims to still have access to Anthropic's Claude Fable 5 model after its global suspension by the US government. He jokingly offered to build startups for commenters, highlighting the ongoing demand for the model's agentic capabilities.

OPEN SOURCE59m ago

LMCache is an open-source KV cache management layer that supercharges LLM inference by sharing and reusing KV caches across GPUs, CPUs, and local/remote storage layers.

LMCache optimizes large language model (LLM) inference by extracting the Key-Value (KV) cache from GPU memory and treating it as a persistent, reusable asset rather than temporary, ephemeral data. By storing the KV cache across a tiered storage hierarchy—including CPU RAM, local disks, and remote backends like Redis or S3—LMCache enables prefix reuse across different queries, sessions, and physical machines. This decouples caching from the inference engine itself, offering integrations with popular platforms like vLLM and SGLang to drastically reduce Time-to-First-Token (TTFT) and boost serving throughput.

Moonshot AI releases Kimi K2.7-Code, an open-source model for coding and agentic workflows that uses 30% fewer reasoning tokens while delivering double-digit benchmark improvements.