Kimi K2.5
Kimi K2.5 was released in January 2026 and remains a strong open-weight option for long-context coding and reasoning, scoring 76.8% on SWE-Bench Verified.
Moonshot · Kimi
Kimi K2.5
Previous-generation Kimi K2 flagship
Kimi K2.5 was released in January 2026 and remains a strong open-weight option for long-context coding and reasoning, scoring 76.8% on SWE-Bench Verified.
Key features
Kimi K2.5 · AI Models
Context window
256K tokens
Released
January 2026
Pricing
Open weights · Kimi API
Key features
- Strong long-context coding and reasoning.
- 76.8% on SWE-Bench Verified.
- Open weights on Hugging Face.
- Stable, production-tested base for fine-tuning.
Best for
Use Kimi K2.5 if you want a stable, production-tested open-weight Moonshot model and don't need K2.6's 12-hour task horizons.
Frequently Asked Questions
Is Kimi K2.5 still supported?
Yes — Moonshot AI maintains K2.5 as a stable production model. The open weights remain on Hugging Face and are actively used by the research and enterprise community.
What score does Kimi K2.5 achieve on SWE-bench?
Kimi K2.5 achieves 76.8% on SWE-Bench Verified, making it one of the highest-scoring open-weight coding models prior to the K2.6 release.
Open Chat
Kimi K2.5 was released in January 2026 and remains a strong open-weight option for long-context coding and reasoning, scoring 76.8% on SWE-Bench Verified.