Kimi K2 Thinking
Moonshot AIhuggingface.co/moonshotai/Kimi-K2-Thinking
Kimi K2 Thinking is Moonshot AI’s most advanced open reasoning model to date, extending the K2 series into agentic, long-horizon reasoning. Built on the trillion-parameter Mixture-of-Experts (MoE) architecture introduced in...
Updated 2 min ago · first seen 11 Sept 2026
model_01M294WW95XPFV3AGY9BV6JMYW
- Parameters
- 1.03T
- T2 · 2 h ago
- Context
- 256K tokens
- T2 · 39 min ago
- Released
- 4 Nov 2025
- T2 · 2 min ago
- License
- other
- T2 · 2 h ago
Specification
- Release date
- 4 Nov 2025
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 min agomedium
- Status
- deprecated
Source:Artificial AnalysisT2observed 39 min agomedium
- Openness
- open-weights
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- License
- other
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Architecture
- DeepseekV3ForCausalLM
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Parameters
- 1.03T
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Context window
- 256K tokens
Source:Artificial AnalysisT2observed 39 min agomedium
- Max output
- 100.3K tokens
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Modalities
- text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Input modalities
- text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Output modalities
- text
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- File size
- 594.3 GB
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Hugging Face repo
- moonshotai/Kimi-K2-Thinking
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Pipeline tag
- text-generation
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Model card
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Downloads
- 51,094
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 min agomedium
- Likes
- 1,713
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 min agomedium
- Gated
- No
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Hf inference providers
- featherless-ai, novita
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Last modified
- 2026-01-30T04:49:39+00:00
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Library name
- transformers
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- License name
- modified-mit
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 54 min agomedium
- Aa median output tokens per second
- 115.2
Source:Artificial AnalysisT2observed 39 min agomedium
- Downloads all time
- 1,875,472
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 min agomedium
- Model type
- kimi_k2
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Openrouter id
- moonshotai/kimi-k2-thinking
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Supported parameters
- frequency_penalty, include_reasoning, logprobs, max_tokens, presence_penalty, reasoning, repetition_penalty, response_format, seed, stop, structured_outputs, temperature, tool_choice, tools, top_k, top_logprobs, top_p
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
- Tags
- transformers, safetensors, kimi_k2, text-generation, custom_code, compressed-tensors
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
- Weights dtype
- BF16, F32, I32
Source:Hugging Face Hub (public pages, model cards, papers)T2observed 2 h agomedium
Each value shows its source, tier and observation time. Conflicting claims are kept side by side and flagged — never averaged. How AI Atlas records facts →
Provenance
Attributed facts
33
Source tiers
T233
Freshest observation
2 min ago
Conflicts
None
Modalities
- Modalities
- text
- Input
- text
- Output
- text
Capabilities
Tool calling
Yes
OpenRouter public model & pricing listing · T2
Structured output
Yes
OpenRouter public model & pricing listing · T2
Reasoning
Yes
OpenRouter public model & pricing listing · T2
Vision
Unavailable
Audio
Unavailable
Fine-tuning available
Unavailable
- Context window
- 256K tokens
Source:Artificial AnalysisT2observed 39 min agomedium
- Max output
- 100.3K tokens
Source:OpenRouter public model & pricing listingT2observed 2 h agomedium
| Benchmark | Score | Metric | Config | Evaluated | Source |
|---|---|---|---|---|---|
| τ²-bench | 92.98% | pass^1 | aa_slug=kimi-k2-thinking · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Artificial Analysis Intelligence Index | 22.02 | index | aa_slug=kimi-k2-thinking · version=4.3 · estimated=true | — | artificialanalysis.aiT2 |
| GPQA | 83.84% | accuracy | aa_slug=kimi-k2-thinking · variant=GPQA Diamond · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| Humanity's Last Exam | 23.82% | accuracy | aa_slug=kimi-k2-thinking · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| IFBench | 68.1% | accuracy | aa_slug=kimi-k2-thinking · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
| SWE-bench Verified | 63.4% | resolved | date=2025-12-10 · board=Verified · system=mini-SWE-agent · model_tag=Kimi-K2-Thinking | 10 Dec 2025 | swebench.comT2 |
| Terminal-Bench | 31.06% | accuracy | aa_slug=kimi-k2-thinking · variant=hard · evaluator=Artificial Analysis · index_version=4.3 | — | artificialanalysis.aiT2 |
Scores are reported as published, with their evaluation configuration (harness, prompting, judge). Results with different configs are not directly comparable — see methodology.
Current prices
| Provider | Input / 1M | Output / 1M | Cached in | Batch in / out | Context | Observed | Source |
|---|---|---|---|---|---|---|---|
| Moonshot AI Platformmoonshotai/kimi-k2-thinking | $0.60 | $2.5 | $0.15 | — | 262.1K | 44 min ago | openrouter.aiT2 |
USD per 1M tokens as published by each provider (USD). Rows are append-only: every change is kept in the history below.
Available through 1
Price history
Input price · USD / 1M tokens
Output price · USD / 1M tokens
- Moonshot AI Platform
Memory need = bytes per parameter (4-bit ≈ 0.5 × 1.15 overhead, 8-bit 1.0, fp16 2.0) + a KV-cache allowance. Not a measurement.
| Hardware | Quantization | Memory | Est. need | Fits |
|---|---|---|---|---|
| NVIDIA DGX B200 | 4bit | 1,440 GB | 590.7 GB | Yes |
| Apple M3 Ultra | 4bit | — | 590.7 GB | No |
| Mac Studio (Apple M5 Ultra) | 4bit | — | 590.7 GB | No |
| AMD Instinct MI325X | 4bit | 256 GB | 590.7 GB | No |
| AMD Instinct MI300X | 4bit | 192 GB | 590.7 GB | No |
| Apple M2 Ultra | 4bit | — | 590.7 GB | No |
| NVIDIA B200 | 4bit | 180 GB | 590.7 GB | No |
| NVIDIA H200 | 4bit | 141 GB | 590.7 GB | No |
| NVIDIA H200 NVL | 4bit | 141 GB | 590.7 GB | No |
| Apple M1 Ultra | 4bit | — | 590.7 GB | No |
| Apple M3 Max | 4bit | — | 590.7 GB | No |
| Apple M4 Max | 4bit | — | 590.7 GB | No |
| Mac Studio (Apple M5 Max) | 4bit | — | 590.7 GB | No |
| MacBook Pro (Apple M5 Max) | 4bit | — | 590.7 GB | No |
| NVIDIA DGX Spark | 4bit | 128 GB | 590.7 GB | No |
| Apple M2 Max | 4bit | — | 590.7 GB | No |
| NVIDIA H100 NVL | 4bit | 94 GB | 590.7 GB | No |
| NVIDIA A100 80GB | 4bit | 80 GB | 590.7 GB | No |
| NVIDIA H100 SXM | 4bit | 80 GB | 590.7 GB | No |
| Apple M1 Max | 4bit | — | 590.7 GB | No |
| Apple M4 Pro | 4bit | — | 590.7 GB | No |
| Mac mini (Apple M5 Pro) | 4bit | — | 590.7 GB | No |
| MacBook Pro (Apple M5 Pro) | 4bit | — | 590.7 GB | No |
| Apple M3 Pro | 4bit | — | 590.7 GB | No |
| Apple M1 Pro | 4bit | — | 590.7 GB | No |
| Apple M2 Pro | 4bit | — | 590.7 GB | No |
| Apple M4 | 4bit | — | 590.7 GB | No |
| iMac (Apple M4) | 4bit | — | 590.7 GB | No |
| Mac mini (Apple M6) | 4bit | — | 590.7 GB | No |
| MacBook Air (Apple M5) | 4bit | — | 590.7 GB | No |
| MacBook Pro (Apple M5) | 4bit | — | 590.7 GB | No |
| NVIDIA GeForce RTX 5090 | 4bit | 32 GB | 590.7 GB | No |
| Apple M2 | 4bit | — | 590.7 GB | No |
| Apple M3 | 4bit | — | 590.7 GB | No |
| NVIDIA GeForce RTX 3090 | 4bit | 24 GB | 590.7 GB | No |
| NVIDIA GeForce RTX 4090 | 4bit | 24 GB | 590.7 GB | No |
| Apple M1 | 4bit | — | 590.7 GB | No |
No lineage recorded
Papers 0
No papers linked yet.
Repositories 0
No repositories linked yet.
As of
Rewind the record: see this entity's attributes exactly as AI Atlas knew them on a given day.
Claim history · License name
License namelicense_name1
| Value | Valid from → to | Status | Source | Confidence | Extractor |
|---|---|---|---|---|---|
| modified-mit | → current | current | Hugging Face Hub (public pages, model cards, papers)T2 | medium | deterministic |
Claims are temporal and append-only: a new observation closes the previous claim (valid_to) instead of overwriting it. Conflicting claims from different sources are kept side by side and flagged — never averaged. Methodology →
Kimi K2 Thinking scores 63.4% on SWE-bench Verified
swebench_leaderboardKimi K2 Thinking scores 31.06% on Terminal-Bench
artificial_analysisKimi K2 Thinking scores 92.98% on τ²-bench
artificial_analysisKimi K2 Thinking scores 68.1% on IFBench
artificial_analysisKimi K2 Thinking scores 23.82% on Humanity's Last Exam
artificial_analysisKimi K2 Thinking scores 83.84% on GPQA
artificial_analysisKimi K2 Thinking scores 22.02 on Artificial Analysis Intelligence Index
artificial_analysisKimi K2 Thinking: context length changed from 262144 to 256000
Context window262.1K tokens→256K tokensartificial_analysisKimi K2 Thinking: release date changed from 2025-11-04 to 2025-11-06
Release date4 Nov 2025→6 Nov 2025openrouterKimi K2 Thinking: release date changed from 2025-11-06 to 2025-11-04
Release date6 Nov 2025→4 Nov 2025huggingfaceMoonshot AI Platform lists Kimi K2 Thinking at $0.6 in / $2.5 out per 1M tokens
openrouter
| Source | Document | Type | Tier | Last observed | Snapshots |
|---|---|---|---|---|---|
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/moonshotai/Kimi-K2-Thinking | model_page | T2· Quality secondary | 2 min ago | 3 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/models?author=moonshotai&p=0&sort=downloads | listing | T2· Quality secondary | 28 min ago | 3 |
| SWE-bench leaderboards | swebench.com/ | leaderboard | T2· Quality secondary | 34 min ago | 1 |
| Artificial Analysis | artificialanalysis.ai/leaderboards/models | leaderboard | T2· Quality secondary | 39 min ago | 1 |
| OpenRouter public model & pricing listing | openrouter.ai/api/v1/models | catalogue | T2· Quality secondary | 44 min ago | 2 |
| Hugging Face Hub (public pages, model cards, papers) | huggingface.co/moonshotai/Kimi-K2-Thinking/raw/main/README.md | model_card | T2· Quality secondary | 53 min ago | 1 |
Tier 1 = official/primary, 2 = quality secondary, 3 = community, 4 = unverified. Every snapshot is archived; see all sources and the methodology.