AI Model Release Tracker

Kimi K2 (0905)Open Weight

Released

Kimi K2 (0905) is an AI model released by Moonshot AI on Friday, Sep 5 2025, 56 days after Kimi K2. It is an open-weight model — the trained weights are available to download and run. It is a 1T parameterThe learned values that shape how a model responds. B means billion and T means trillion. For mixture-of-experts models, the total count is larger than the number active for each token. Parameter count alone does not measure quality. model with a 256k token context windowHow much information the model can work with in one request, measured in tokens — pieces of words or other data. This includes your question, attachments, conversation and reply. Limits can vary by model variant and provider.. Benchmark results (shown below) cover SWE-Bench Verified, and GPQA Diamond.

Available from

Novita$0.60$2.50256Kfp8
USD per 1M tokensEvery price here is for one million tokens. A token is roughly three-quarters of a word, so a million tokens is about 750,000 words of text.Checked November 3, 2026Third-party rates from openrouter.ai, not the lab’s own list price.

Benchmarks

Coding

SWE-Bench VerifiedCoding — Real coding tasks pulled from open-source projects — the AI has to find and fix actual bugs. A human-checked version of the original SWE-Bench. Higher is better.
65.8%
#46 of 70

Reasoning & science

GPQA DiamondScience — Graduate-level science questions in biology, physics, and chemistry — hard enough that subject-matter PhDs score around 65%. Higher is better.
75.1%
#48 of 67

Compare Kimi K2 (0905) with

Kimi K2 (0905)

Suggested comparisons

Kimi K2 (0905)vsKimi K3.1 ProKimi K2 (0905)vsGPT-6.1 TerraKimi K2 (0905)vsClaude Haiku 5.5Kimi K2 (0905)vsGemini 4 CarbonKimi K2 (0905)vsMuse Spark OpenKimi K2 (0905)vsGrok 4.7Kimi K2 (0905)vsDeepSeek-V4.1-Flash-1017Kimi K2 (0905)vsMistral Large 4Kimi K2 (0905)vsGLM-5.3-FlashKimi K2 (0905)vsQwen3.8-Max-0902Kimi K2 (0905)vsNemotron 3.5 Bolt E4B 4GB

About

Kimi K2 (0905) was Moonshot AI's September 2025 update to the K2 instruction model. It retained the trillion-parameter mixture-of-experts design with approximately 32 billion active parameters per token, while expanding context from 128,000 to 256,000 tokens. The model card emphasized stronger agentic coding and improvements in both the appearance and practicality of generated frontend code.

The dated checkpoint matters because it was evaluated separately from the original July release. Moonshot reported 69.2% on SWE-bench Verified across five runs under its documented harness, compared with 65.8% for the earlier version. The release offered a more capable instruction checkpoint within the existing K2 generation, before the separately named Thinking model introduced a different reasoning profile.

Frequently asked questions

Kimi K2 (0905) was released by Moonshot AI on Friday, Sep 5 2025.

All Moonshot AI releases

14 tracked

2023

1 release
Kimi Chat
Oct 9 2023