# Kimi K2.5 Benchmark Scores & Performance

> Kimi K2.5 by Moonshot AI scores 53.05/100 overall, ranking #86 out of 954 AI models.

Canonical page: https://benchlm.ai/models/kimi-k2-5

Last updated: October 17, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | Moonshot AI |
| Source Type | Open Weight |
| Reasoning Type | Non-Reasoning |
| Context Window | 256K |
| Overall Score | 53.05/100 |
| Overall Rank | #86 of 954 |

## Family & Coverage

- Family: Kimi K2.5
- Variant: base
- Benchmarks covered: 84 of 639
- Sibling models: [Kimi K2.5 (Reasoning)](/models/kimi-k2-5-reasoning)
- Coverage note: BenchLM currently has partial benchmark coverage for this model, so the overall score is conservative.

## Agentic Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [Terminal-Bench 2.0](/benchmarks/terminal-bench-2) | 50.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [BrowseComp](/benchmarks/browsecomp) | 60.6% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [OSWorld-Verified](/benchmarks/osworld-verified) | 63.3% | [Kimi-K2.6 model card](https://huggingface.co/moonshotai/Kimi-K2.6) |
| [Claw-Eval](/benchmarks/claw-eval) | 52.3% | [Claw-Eval leaderboard](https://claw-eval.github.io/) |
| [QwenClawBench](/benchmarks/qwenclawbench) | 54.3% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [τ³-bench results](/benchmarks/tau3-bench) | 65.7% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [DeepSearchQA](/benchmarks/deepsearchqa) | 77.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [DeepPlanning](/benchmarks/deepplanning) | 14.4% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [Toolathlon](/benchmarks/toolathlon) | 27.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [MCP Atlas](/benchmarks/mcpatlas) | 29.5% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [MCP-Tasks](/benchmarks/mcptasks) | 59.1% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [WideResearch](/benchmarks/wideresearch) | 72.7% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [τ²-bench results](/benchmarks/tau2-bench) | 95.9% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [BFCL v4](/benchmarks/bfcl-v4) | 68.3% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [APEX-Agents-AA](/benchmarks/apexagentsaa) | 11.5% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [Gert Labs](/benchmarks/gertlabs) | 45.88% | [Gert Labs rankings](https://gertlabs.com/rankings) |
| [ResearchClawBench](/benchmarks/researchclawbench) | 14.0% | [ResearchClawBench leaderboard](https://internscience.github.io/ResearchClawBench-Home/) |
| [JobBench](/benchmarks/jobbench) | 8.7% | [JobBench paper](https://arxiv.org/abs/2605.26329) |
| [GDPval-AA](/benchmarks/gdpvalaanormalized) | 17.0% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [GDPval-AA](/benchmarks/gdpvalaa) | 936 | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |

## Coding Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [SWE-bench Verified](/benchmarks/swe-bench-verified) | 76.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [SWE-bench Verified*](/benchmarks/swe-bench-verified) | 70.8% | [Arcee Trinity-Large-Thinking comparison table](https://www.arcee.ai/blog/trinity-large-thinking) |
| [LiveCodeBench v6](/benchmarks/livecodebench-v6) | 85.0% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [SWE-bench Pro](/benchmarks/swe-bench-pro) | 50.7% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [SWE Multilingual](/benchmarks/swe-bench-multilingual) | 73% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [SWE-Rebench](/benchmarks/swe-rebench) | 58.5% | [SWE-Rebench leaderboard](https://swe-rebench.com/?insight=sep_2025) |
| [React Native Evals](/benchmarks/reactnativeevals) | 77.2% | [React Native Evals leaderboard](https://rn-evals.vercel.app/) |
| [NL2Repo](/benchmarks/nl2repo) | 32% | [GLM-5.1 model card comparison table](https://huggingface.co/zai-org/GLM-5.1) |
| [SciCode](/benchmarks/scicode) | 48.7% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [Terminal-Bench 2.0](/benchmarks/terminal-bench-2) | 50.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [AA Coding Index](/benchmarks/aacodingindex) | 46.8% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |

## Multimodal & Grounded Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [MMMU](/benchmarks/mmmu) | 84.3% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [MMMU-Pro](/benchmarks/mmmu-pro) | 78.5% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [RealWorldQA](/benchmarks/realworldqa) | 81.0% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [OmniDocBench 1.5](/benchmarks/omnidocbench15) | 88.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [Video-MME](/benchmarks/videomme) | 87.4% | [Reported upstream source](https://www.kimi.com/blog/kimi-k2-5.html) |
| [Video-MME (with subtitle)](/benchmarks/videommewithsub) | 87.4% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [Video-MME (w/o subtitle)](/benchmarks/videommenosub) | 83.2% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [MathVision](/benchmarks/mathvision) | 84.2% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [We-Math](/benchmarks/wemath) | 84.7% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [DynaMath](/benchmarks/dynamath) | 84.4% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [SimpleVQA](/benchmarks/simplevqa) | 71.2% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [MMLongBench-Doc](/benchmarks/mmlongbenchdoc) | 58.5% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [CC-OCR](/benchmarks/ccocr) | 79.7% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [AI2D_TEST](/benchmarks/ai2dtest) | 90.8% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [CountBench](/benchmarks/countbench) | 94.1% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [RefCOCO (avg)](/benchmarks/refcocoavg) | 87.8% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [MMVU](/benchmarks/mmvu) | 80.4% | [Reported upstream source](https://www.kimi.com/blog/kimi-k2-5.html) |
| [VideoMMMU](/benchmarks/videommmu) | 86.6% | [Qwen3.6-Plus multimodal comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [MLVU (M-Avg)](/benchmarks/mlvuavg) | 85.0% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [CharXiv](/benchmarks/charxiv) | 77.5% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [AA-MMMU-Pro](/benchmarks/mmmu-pro) | 75.4% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [Design Arena Website](/benchmarks/designarenawebsite) | 1250 | [OpenRouter model benchmarks](https://openrouter.ai/moonshotai/kimi-k2.5/benchmarks) |

## Reasoning Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [LongBench v2](/benchmarks/longbench-v2) | 61% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [AA-LCR](/benchmarks/lcr) | 78.0% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [CritPt](/benchmarks/critpt) | 3.1% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |

## Knowledge Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [GPQA](/benchmarks/gpqa) | 87.6% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [GPQA-D](/benchmarks/gpqa-diamond) | 87.6% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [SuperGPQA](/benchmarks/supergpqa) | 69.2% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [MMLU-Pro](/benchmarks/mmlu-pro) | 87.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [MMLU-Pro (Arcee)](/benchmarks/mmlu-pro) | 87.1% | [Arcee Trinity-Large-Thinking comparison table](https://www.arcee.ai/blog/trinity-large-thinking) |
| [MMLU-Redux](/benchmarks/mmluredux) | 94.5% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [C-Eval](/benchmarks/ceval) | 94% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [HLE](/benchmarks/hle) | 30.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [Artificial Analysis Intelligence Index](/benchmarks/artificialanalysis) | 23.5% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [AA-GPQA Diamond](/benchmarks/gpqa-diamond) | 87.9% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [AA-HLE](/benchmarks/aahle) | 30.7% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [AA-Omniscience Index](/benchmarks/aaomniscienceindex) | -7.3% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [AA-Omniscience Accuracy](/benchmarks/omniscienceaccuracy) | 35.2% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |
| [AA-Omniscience Hallucination Rate](/benchmarks/omnisciencehallucinationrate) | 65.7% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |

## Instruction Following Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [IFEval](/benchmarks/ifeval) | 93.9% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [AA-IFBench](/benchmarks/ifbench) | 70.2% | [Artificial Analysis model benchmarks](https://artificialanalysis.ai/models/kimi-k2-5) |

## Multilingual Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [MMLU-ProX](/benchmarks/mmluprox) | 82.3% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [NOVA-63](/benchmarks/nova63) | 56.0% | [Qwen3.6-Plus comparison table](https://qwen.ai/blog?id=qwen3.6) |
| [INCLUDE](/benchmarks/include) | 83.3% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [PolyMath](/benchmarks/polymath) | 43.1% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |
| [MAXIFE](/benchmarks/maxife) | 72.8% | [Qwen3.5-397B-A17B model card comparison table](https://huggingface.co/Qwen/Qwen3.5-397B-A17B) |

## Mathematics Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [AIME 2025](/benchmarks/aime2025) | 96.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [AIME26](/benchmarks/aime2026) | 95.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [AIME25 (Arcee)](/benchmarks/aime2025arcee) | 96.3% | [Arcee Trinity-Large-Thinking comparison table](https://www.arcee.ai/blog/trinity-large-thinking) |
| [HMMT Feb 2025](/benchmarks/hmmtfeb2025) | 95.4% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [HMMT Nov 2025](/benchmarks/hmmtnov2025) | 91.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [HMMT Feb 2026](/benchmarks/hmmtfeb2026) | 87.1% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |
| [MMAnswerBench](/benchmarks/mmanswerbench) | 81.8% | [MoonshotAI: Kimi K2.5 model card](https://huggingface.co/moonshotai/Kimi-K2.5) |

## Other Moonshot AI Models

- [Kimi K3](/models/kimi-k3) - Score: 69.87
- [Kimi K2.6](/models/kimi-2-6) - Score: 58.4
- [Kimi K2.5 (Reasoning)](/models/kimi-k2-5-reasoning) - Score: 56.09
- [Kimi K2.7 Code](/models/kimi-k2-7-code) - Score: 49.52
- [Kimi K2](/models/kimi-k2) - Score: 33.48
- [Moonshot v1](/models/moonshot-v1) - Score: not computed
- [Kimi-Audio 7B](/models/kimi-audio-7b) - Score: not computed