Kimi K2 Thinking
Kimi K2 Thinking is an open-weights language model from Moonshot AI, released November 6, 2025. It has 1T parameters (mixture-of-experts, 33B active). The context window is 256K tokens. On the Epoch Capabilities Index it scores 145.4, rank 93 of 256 rated models.
Today, we are introducing Kimi K2 Thinking, our best open-source thinking model. Built as a thinking agent, it reasons step by step while using tools, achieving state-of-the-art performance on Humanity's Last Exam (HLE), BrowseComp, and other benchmarks, with major gains in reasoning, agentic search, coding, writing, …
Check it against your own PC →
Can you run Kimi K2 Thinking locally?
Even with 32 GB of VRAM and 32 GB of RAM it is too large to run at a usable speed on a single consumer GPU. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Too large | – | – | – | – |
| 8 GB VRAM | Too large | – | – | – | – |
| 12 GB VRAM | Too large | – | – | – | – |
| 16 GB VRAM | Too large | – | – | – | – |
| 24 GB VRAM | Too large | – | – | – | – |
| 32 GB VRAM | Too large | – | – | – | – |
Ready-made quantizations on Hugging Face: unsloth/Kimi-K2-Thinking-GGUF · bartowski/moonshotai_Kimi-K2-Thinking-GGUF · DevQuasar/moonshotai.Kimi-K2-Thinking-GGUF · ubergarm/Kimi-K2-Thinking-GGUF · squ11z1/Hypnos-Colossus-1T · nvidia/Kimi-K2-Thinking-NVFP4
Strengths by category
ModelTape's 0–100 fit on Epoch AI benchmark results.
| Category | Score | Based on |
|---|---|---|
| Math | 49.0 | OTIS Mock AIME 2024-2025, FrontierMath-2025-02-28-Private, FrontierMath-Tier-4-2025-07-01-Private |
Benchmark results
| Benchmark (Epoch AI) | Score |
|---|---|
| OTIS Mock AIME | 83.0% |
| GPQA Diamond | 82.2% |
| FrontierMath-2025-02-28-Private | 37.6% |
| Terminal-Bench | 35.7% |
| SimpleBench | 27.5% |
| FrontierMath-Tier-4-2025-07-01-Private | 0.0% |
Links
Announcement · Hugging Face: moonshotai/Kimi-K2-Thinking · Epoch AI benchmarks
Models with a similar score
GPT-5.4 Mini · Qwen 3.6 Plus · Kimi K2.5 · Inkling · Qwen3.5 397B-A17B · Qwen3.7-Plus · MiniMax-M3 · Claude Sonnet 4.5
More from Moonshot AI
Kimi K3.1 Pro · Kimi K3 · Kimi K2.7 Code · Kimi K2.6 · Kimi K2.5 · Kimi Linear · Kimi K2 (Jul 2025) · Kimi-VL