Moonshot AI model product
Kimi K2 Thinking
Kimi K2 Thinking is an open-source language model from Moonshot AI designed for step-by-step reasoning and dynamic tool invocation.
Updated Oct 23, 2026. Default version: Kimi K2-Thinking-0905
LLMBoard Score58.8Kimi K2-Thinking-0905
Coverage100%20 benchmark families
Context window262.1KTokens
Official input priceN/AOfficial price unavailable
Kimi K2 Thinking Capability Profile
This profile uses the model's current scored version. Arena ratings and prices are shown separately.
Kimi K2 Thinking Benchmark Results
Benchmark scores for Kimi K2-Thinking-0905.
Sort by
Benchmark | Score | Rank | Participants | Percentile | Evidence | Evaluated |
|---|
| BenchmarkFinSearchComp-T3 | Score47.40% | Rank01 | Participants1 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkFRAMES | Score87.00% | Rank01 | Participants3 | Percentile100.00% | EvidenceC | Evaluated |
| BenchmarkOJBench | Score48.70% | Rank02 | Participants9 | Percentile87.50% | EvidenceC | Evaluated |
| BenchmarkSeal-0 | Score56.30% | Rank02 | Participants7 | Percentile83.33% | EvidenceC | Evaluated |
| BenchmarkTerminal-Bench | Score47.10% | Rank03 | Participants25 | Percentile91.67% | EvidenceC | Evaluated |
| BenchmarkHMMT 2025 | Score97.50% | Rank04 | Participants33 | Percentile90.63% | EvidenceC | Evaluated |
| BenchmarkAIME 2025 | Score100.00% | Rank05 | Participants122 | Percentile96.69% | EvidenceC | Evaluated |
| BenchmarkMMLU-Redux | Score94.40% | Rank05 | Participants50 | Percentile91.84% | EvidenceC | Evaluated |
| BenchmarkMulti-SWE-Bench | Score41.90% | Rank05 | Participants7 | Percentile33.33% | EvidenceC | Evaluated |
| BenchmarkHealthBench | Score58.00% | Rank06 | Participants14 | Percentile61.54% | EvidenceC | Evaluated |
| BenchmarkAA AIME 2025 | Score94.67% | Rank08 | Participants96 | Percentile92.63% | EvidenceB | EvaluatedN/A |
| BenchmarkBrowseComp-zh | Score62.30% | Rank09 | Participants15 | Percentile42.86% | EvidenceC | Evaluated |
| BenchmarkSciCode | Score44.80% | Rank14 | Participants31 | Percentile56.67% | EvidenceC | Evaluated |
| BenchmarkWritingBench | Score73.80% | Rank16 | Participants16 | Percentile0.00% | EvidenceC | Evaluated |
| BenchmarkLiveCodeBench v6 | Score83.10% | Rank19 | Participants67 | Percentile72.73% | EvidenceC | Evaluated |
| BenchmarkIMO-AnswerBench | Score78.60% | Rank20 | Participants22 | Percentile9.52% | EvidenceC | Evaluated |
| BenchmarkHumanity's Last Exam | Score51.00% | Rank22 | Participants109 | Percentile80.56% | EvidenceC | Evaluated |
| BenchmarkMMLU-Pro | Score84.60% | Rank29 | Participants142 | Percentile80.14% | EvidenceC | Evaluated |
| BenchmarkSWE-bench Multilingual | Score61.10% | Rank38 | Participants50 | Percentile24.49% | EvidenceC | Evaluated |
| BenchmarkBrowseComp | Score60.20% | Rank44 | Participants70 | Percentile37.68% | EvidenceC | Evaluated |
| BenchmarkAA IFBench | Score68.10% | Rank55 | Participants168 | Percentile67.66% | EvidenceB | EvaluatedN/A |
| BenchmarkAA Omniscience Accuracy | Score30.87% | Rank58 | Participants201 | Percentile71.50% | EvidenceB | EvaluatedN/A |
| BenchmarkSWE-Bench Verified | Score71.30% | Rank63 | Participants120 | Percentile47.90% | EvidenceC | Evaluated |
| BenchmarkAA CritPt | Score2.57% | Rank69 | Participants200 | Percentile65.83% | EvidenceB | EvaluatedN/A |
| BenchmarkGPQA | Score84.50% | Rank73 | Participants256 | Percentile71.76% | EvidenceC | Evaluated |
| BenchmarkAA HLE Text No Tools | Score23.82% | Rank73 | Participants200 | Percentile63.82% | EvidenceB | EvaluatedN/A |
| BenchmarkAA GPQA Diamond | Score83.84% | Rank80 | Participants199 | Percentile60.10% | EvidenceB | EvaluatedN/A |
| BenchmarkAA LCR v1.1 | Score72.00% | Rank82 | Participants193 | Percentile57.81% | EvidenceB | EvaluatedN/A |
Kimi K2 Thinking Arena Results
Preference and agent-evaluation results for the default version.
No Arena results
The default version does not have a matching Arena result yet.
Kimi K2 Thinking Pricing
Official vendor API pricing appears first, followed by individual provider offers.
- Official API
- N/A
- Official provider
- N/A
- Lowest third-party
- From $0.40 input, $2.5 output per 1M via OpenCode
- Tracked offerings
- 4
Sort by
Provider | Provider model ID | Region | Input / 1M | Output / 1M | Context | Updated |
|---|
| ProviderOpenCode | Provider model IDkimi-k2-thinking | Regionglobal | Input / 1M$0.40 | Output / 1M$2.5 | Context262.1K | Updated |
| ProviderHelicone | Provider model IDkimi-k2-thinking | Regionglobal | Input / 1M$0.48 | Output / 1M$2 | Context256K | Updated |
| Provider302.AI | Provider model IDkimi-k2-thinking | Regionglobal | Input / 1M$0.575 | Output / 1M$2.3 | Context262.1K | Updated |
| ProviderCharm Hyper | Provider model IDkimi-k2-thinking | Regionglobal | Input / 1M$0.60 | Output / 1M$2.5 | Context262.1K | Updated |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
Kimi K2 Thinking Runtime Performance
Provider-specific output speed and catalog latency for Kimi K2-Thinking-0905. Runtime does not affect the capability score.
No runtime data
No provider-specific speed or latency record is linked to the default version yet.
Output Speed is generated output tokens received per second. Catalog latency is reported separately from observed provider TTFT.
Kimi K2 Thinking Specifications
Technical details for the model's default version.
- Version
- Kimi K2-Thinking-0905
- Released
- Sep 5, 2025
- Knowledge cutoff
- Unknown
- Parameters
- 1T
- Context window
- 262.1K
- Max output
- 262.1K
- Inputs
- text
- Outputs
- text
- Open weights
- Yes
- License
- MIT
Kimi K2 Thinking Versions
Available versions of this model. The score column identifies the version used in the overall ranking.
Sort by
Version | Released | LLMBoard | Parameters | Context | Max output | Open weights | License |
|---|
| VersionKimi K2-Thinking-0905 | Released | LLMBoard58.77 | Parameters1T | Context262.1K | Max output262.1K | Open weightsYes | LicenseMIT |
Models similar to Kimi K2 Thinking
Recommendations prioritize the same model type and family, then the closest LLMBoard score.
MA
Kimi K2.7 Code
Moonshot AI
MA
Kimi K2.5
Moonshot AI
MA
Kimi K2.6
Moonshot AI
MA
Kimi K2
Moonshot AI
MA
Kimi K3
Moonshot AI
MA
Kimi k1.5
Moonshot AI
What is Kimi K2 Thinking?
Key information about Kimi K2 Thinking and its available data.
Kimi K2 Thinking is a thinking model trained to interleave chain-of-thought reasoning with function calls. It uses native INT4 quantization via Quantization-Aware Training (QAT), has a 256k context window, and supports up to 200–300 consecutive tool invocations.
Data as of 2026-10-22.
FAQ
Common questions about Kimi K2 Thinking.
When was Kimi K2 Thinking released?
Kimi K2 Thinking's default version was released on Sep 5, 2025.
How much does Kimi K2 Thinking cost?
No official standard PAYG price is currently available for Kimi K2 Thinking. The lowest tracked third-party offer starts at $0.40 input and $2.5 output via OpenCode.
Who created Kimi K2 Thinking?
Kimi K2 Thinking was created by Moonshot AI.
What is the context window for Kimi K2 Thinking?
The default version has a 262.1K token context window.
Is Kimi K2 Thinking open weight?
Yes. The default version is marked as open weight under MIT.
How many API providers offer Kimi K2 Thinking?
4 provider offerings are linked to the default version.