/Tech1h ago

Moonshot AI open-sources Kimi-K2.7-Code, cutting reasoning tokens by 30% and boosting Kimi Code Bench scores by 21.8%

Story Overview

Moonshot AI has positioned Kimi K2.7 Code as its latest coding-focused release, with API access now available on the official platform and an emphasis on steadier performance across extended tasks plus clearer instruction handling. While the headline flags open-sourcing and concrete benchmark lifts, official materials do not yet confirm weight releases or those exact deltas, leaving the scope of public availability and measured gains to be clarified.

5624.7K80769212.4K

#501

Original post

Kimi.ai@Kimi_Moonshot

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

🔷 Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite. 🔷 Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. 🔷 Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates.

⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code.

🔗 Kimi Code: https://kimi.com/code 🔗 API: https://platform.moonshot.ai

3:16 AM · Jun 12, 2026 · 224K Views

Pricing Watch

Token Rates Set for Coders

Input tokens run $0.95 per million, outputs $4.00 per million, and cache hits drop to $0.19 per million, with a limited-time promotion noted on the platform.

Open Question

Long-Horizon Claims Await Checks

The model highlights better stability on lengthy coding jobs and instruction following, yet independent confirmation of the 30 % token cut or 21.8 % bench jump remains absent from current sources.

Sentiment

Many users congratulated Moonshot AI for open-sourcing Kimi-K2.7-Code because of its major coding gains, efficiency, and benchmarks, while some dismissed the model as outdated or average.

Pos

77.4%

Neg

22.6%

92 comments with sentiment.

Cluster Engagement

Posts from X

Most Activity

VIEWS9.5K

Zephyr@zephyr_z9

Great work from KIMI

Kimi.ai@Kimi_Moonshot

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code.

🔗 Kimi Code: https://kimi.com/code 🔗 API: https://platform.moonshot.ai

1h9.5K644

BOOKMARKS12LIKES108REPLIES9

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

I want to see this compared with Composer 2.5 Like, really hard Cursor has a ton of proprietary data, a large head start, and threw a Colossus at RLing Kimi K2.5 checkpoint. What is the gap now?

Kimi.ai@Kimi_Moonshot

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code.

🔗 Kimi Code: https://kimi.com/code 🔗 API: https://platform.moonshot.ai

1h4.6K10812

RETWEETS1

Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex

HUGE @htihle and others will get to test this soon I hope

Young@Young_AGI

Kimi K2.7 Code 发布并开源

1h2.3K425

Kimi.ai@Kimi_Moonshot

🎸 We're also launching the Kimi Code Beta Program today. Apply now if you'd like to try upcoming models and features before public release 👉 https://www.kimi.com/code/beta

1h1.8K344

Cody Blakeney@code_star

It would be really cool if the top Chinese labs could properly benchmark and evaluate Fable/Mythos to show how it compares to their releases. I guess that isn’t possible though.

Kimi.ai@Kimi_Moonshot

🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code.

🔗 Kimi Code: https://kimi.com/code 🔗 API: https://platform.moonshot.ai

44m1.1K332

Kimi.ai@Kimi_Moonshot

🔗 Weights & code: https://huggingface.co/moonshotai/Kimi-K2.7-Code

1h923163

Ahmad@TheAhmadOsman

@code_star You want another “omgggg china is distilling us ban opensource” post 🤔😂

42m13081

__@ooga333

@TobiasVinc57321 @Kimi_Moonshot It's literally 99% the same wtf are you crying about

1h1306

Julien@julienzeroshot

@Kimi_Moonshot Will this be comparable in price per token with the base Kimi 2.7?

1h1.5K3

Hamza@thegenioo

@Kimi_Moonshot The best I read here is "Less overthinking" cuz 2.6 wastes so much time on that

Looking forward to testing this, kudos to the team

1h4338

Hamza@thegenioo

@teortaxesTex It is more expensive than composer 2.5 tho

50m914

Mohammad Ayaan@moayaan1911

@Kimi_Moonshot Hope its added soon on opencode go and commandcode

1h6314

bart moc@bmocior

@Kimi_Moonshot @opencode when?

1h1711

tejas@persuingdreams

@Kimi_Moonshot I know it's a dumb question but, when I run the locally downloaded coding model via opencode it takes a hell lot of time even for small code fixes. Has this issue tackled in this version?

1h417

cedric@cedric_chee

@Kimi_Moonshot API Pricing

1h912

dr corn farmer, phd 👨‍⚕️ 🌽 👩‍🌾@farmerofcorn

@TheLouieCo @Kimi_Moonshot plz quit your job as “ai consultant”

1h755

cheaty@cheatyyyy

@thegenioo @teortaxesTex because cursor has massive infra deals with fireworks to host these models for cheap, but as far as anyone knows its still 1T-32b active, so should be identical cost to run

49m192

Matt@MatthewEuc

@Kimi_Moonshot OMG

1h166

Canopy Wave@CanopyWave_AI

@Kimi_Moonshot Kimi-K2.7-Code looks insane. Less overthinking + better long-horizon performance = actual usable coding agent. Congrats on the release!

1h3213

Zaid@zaidmukaddam

@Kimi_Moonshot insane shippp

1h3093