Kimi K2
Moonshot AI ·
Verified
3 of 3 published values traced to a source2 primary sources, 2 archived
Why this status
Every value this record asserts — date (2025-07-11), context window (128K) and parameter count (1T) — was found in an archived primary source, checked 2026-08-10. Modalities recorded as text-to-text. The primary source states the model is text-only. Capabilities evidenced from the primary sources: coding: “Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for ag” | agentic: “Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for ag” | reasoning: “Evaluation Results Instruction model evaluation results Benchmark Metric Kimi K2 Instruct DeepSeek-V3-0324 Qwen3-235B-A22B (non-thinking) Claude Sonnet 4 (w/o extended th”.
What these states mean · How records are judged · Has this record changed?
A trillion-parameter open MoE with 32B active parameters excelling at code and agents.
Details
| Released | July 11, 2025stated in huggingface.coKimi K2 was released on July 11, 2025.
|
|---|---|
| Company | Moonshot AI |
| Family | Kimi |
| Lineage | before Kimi K3 |
| Era | Reasoning models |
| Type | Language model · LLM |
| Modalities | Text → Text |
| Capabilities | Agentic, Coding, Reasoning |
| Weights | Open weights |
| Licence | Modified MIT License |
| Context window | 128K tokensstated in huggingface.coKimi K2 supports a context window of 128,000 tokens.
|
| Parameters | 1Tstated in huggingface.coKimi K2 has 1,000,000,000,000 parameters.
|
Current information, elsewhere
This record is a historical one: every figure above is traced to a dated primary source. For what is true today — live pricing, current rankings, availability — these are better maintained than anything here could be.
- Current price and providers OpenRouter ↗
- Current benchmark scores Artificial Analysis ↗
- Weights and model card Hugging Face ↗
Sources
Record status: verified · confidence 90/100 · 2 of 2 primary
Every value this record asserts — date (2025-07-11), context window (128K) and parameter count (1T) — was found in an archived primary source, checked 2026-08-10. Modalities recorded as text-to-text. The primary source states the model is text-only. Capabilities evidenced from the primary sources: coding: “Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for ag” | agentic: “Trained with the Muon optimizer, Kimi K2 achieves exceptional performance across frontier knowledge, reasoning, and coding tasks while being meticulously optimized for ag” | reasoning: “Evaluation Results Instruction model evaluation results Benchmark Metric Kimi K2 Instruct DeepSeek-V3-0324 Qwen3-235B-A22B (non-thinking) Claude Sonnet 4 (w/o extended th”.
- moonshotai.github.io Official announcement archived
- huggingface.co Model card archived
Primary means published by the organisation that made the model. Verified records require at least one.
Kimi family
- Kimi K2 July 11, 2025
- Kimi K3 July 16, 2026
- Kimi K2.8 Preview September 11, 2026
Compare Kimi K2 with another model →
See Kimi K2 on the interactive timeline →