Moonshot AI model product
Kimi K2.5 is Moonshot AI's flagship agentic model and a new SOTA open model. It unifies vision and text, thinking and non-thinking modes, and single-agent and multi-agent execution into one model. Built with Full-Parameter RL tuning, it achieves state-of-the-art performance across agents, coding, image, and video benchmarks.
Updated Aug 10, 2026. Default version: Kimi K2.5
Structured fields from the published default version.
This profile uses the latest version under this unique model that has a calculated LLMBoard score. Arena and price are excluded.
Published benchmark records for the scored version Kimi K2.5.
| FinSearchComp T2&T3 | 0.7 | 1 | 1 | 100.0% | C | |
| InfoVQAtest | 0.9 | 1 | 12 | 100.0% | C | |
| MathVista-Mini | 0.9 | 1 | 23 | 100.0% | C | |
| MMVU | 0.8 | 1 | 4 | 100.0% | C | |
| OCRBench | 0.9 | 1 | 22 | 100.0% | C | |
| OJBench (C++) | 0.6 | 1 | 1 | 100.0% | C | |
| Seal-0 | 0.6 | 1 | 6 | 100.0% | C | |
| PaperBench | 0.6 | 2 | 3 | 50.0% | C | |
| WideSearch | 0.8 | 2 | 9 | 87.5% | C | |
| AA-LCR | 0.7 | 3 | 15 | 85.7% | C | |
| LongVideoBench | 0.8 | 3 | 4 | 33.3% | C | |
| MotionBench | 0.7 | 3 | 3 | 0.0% | C | |
| VideoMMMU | 0.9 | 3 | 26 | 92.0% | C | |
| ZEROBench | 0.1 | 4 | 9 | 62.5% | C | |
| LVBench | 0.8 | 5 | 24 | 82.6% | C | |
| SimpleVQA | 0.7 | 5 | 13 | 66.7% | C | |
| Video-MME | 0.9 | 5 | 17 | 75.0% | C | |
| WorldVQA | 0.5 | 5 | 5 | 0.0% | C | |
| HMMT 2025 | 1.0 | 6 | 33 | 84.4% | C | |
| LongBench v2 | 0.6 | 6 | 17 | 68.8% | C | |
| DeepSearchQA | 0.8 | 7 | 8 | 14.3% | C | |
| MMLU-Pro | 0.9 | 7 | 129 | 95.3% | C | |
| SciCode | 0.5 | 7 | 18 | 64.7% | C | |
| LiveCodeBench v6 | 0.8 | 9 | 53 | 84.6% | C | |
| OmniDocBench 1.5 | 0.9 | 9 | 16 | 46.7% | C | |
| MathVision | 0.8 | 10 | 32 | 71.0% | C | |
| CyberGym | 0.4 | 11 | 11 | 0.0% | C | |
| IMO-AnswerBench | 0.8 | 13 | 19 | 33.3% | C | |
| SWE-bench Multilingual | 0.7 | 13 | 34 | 63.6% | C | |
| FrontierSWE | 0.3 | 14 | 15 | 7.1% | B | |
| MMMU-Pro | 0.8 | 18 | 65 | 73.4% | C | |
| AIME 2025 | 1.0 | 19 | 114 | 84.1% | C | |
| Humanity's Last Exam | 0.5 | 20 | 92 | 79.1% | C | |
| CharXiv-R | 0.8 | 25 | 47 | 47.8% | C | |
| BrowseComp | 0.7 | 26 | 58 | 56.1% | C | |
| SWE-Bench Verified | 0.8 | 27 | 104 | 74.8% | C | |
| LiveBench | 0.7 | 32 | 38 | 16.2% | B | |
| Terminal-Bench 2.0 | 0.5 | 35 | 49 | 29.2% | C | |
| GPQA | 0.9 | 38 | 233 | 84.0% | C | |
| SWE-Bench Pro | 0.5 | 42 | 44 | 4.7% | C |
Preference and agent-evaluation signals from published Arena datasets.
The default version has no published Arena rows, or its source alias has not been resolved.
Official vendor API PAYG pricing is summarized first. The table then lists individual provider offerings without treating their minimum as the official price.
| Alibaba Coding Plan (China) | kimi-k2.5 | global | N/A | N/A | 262.1K | |
| Alibaba Coding Plan | kimi-k2.5 | global | N/A | N/A | 262.1K | |
| Alibaba Token Plan | kimi-k2.5 | global | N/A | N/A | 262.1K | |
| Tencent Coding Plan (China) | kimi-k2.5 | global | N/A | N/A | 262.1K | |
| Alibaba Token Plan (China) | kimi-k2.5 | global | N/A | N/A | 262.1K | |
| CrofAI | kimi-k2.5 | global | $0.35 | $1.7 | 262.1K | |
| DigitalOcean | kimi-k2.5 | global | $0.375 | $2.03 | 262.1K | |
| LLM Gateway | kimi-k2.5 | global | $0.405 | $1.98 | 262.1K | |
| Cortecs | kimi-k2.5 | global | $0.495 | $2.77 | 262.1K | |
| Auriko | kimi-k2.5 | global | $0.50 | $2.8 | 262.1K | |
| Charm Hyper | kimi-k2.5 | global | $0.55 | $2.89 | 262.1K | |
| Alibaba (China) | kimi-k2.5 | global | $0.574 | $2.41 | 262.1K | |
| OpenCode Go | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| DaoXE | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| OpenCode Zen | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| FrogBot | kimi-k2.5 | global | $0.60 | $3 | 256K | |
| Abacus | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| Azure Cognitive Services | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| Moonshot AI (China) | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| Azure | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| AIHubMix | kimi-k2.5 | global | $0.60 | $3 | 262.1K | |
| Moonshot AI | kimi-k2.5 | global | $0.60 | $3 | 262.1K |
Official prices use only the vendor's configured official Provider and positive standard USD PAYG rates. Third-party offers remain explicitly labeled.
All published versions linked to this unique model. The score columns identify the version used by the current overall ranking.
| Kimi K2.5 | 71.6 | 1T | 262.1K | 262.1K | Yes | MIT |
A concise description based on the published model registry.
Kimi K2.5 is Moonshot AI's flagship agentic model and a new SOTA open model. It unifies vision and text, thinking and non-thinking modes, and single-agent and multi-agent execution into one model. Built with Full-Parameter RL tuning, it achieves state-of-the-art performance across agents, coding, image, and video benchmarks.
Use the benchmark, Arena and pricing sections above as separate evidence. A missing field means the current data snapshot does not support that claim.
Data snapshot: 2026-08-07. Editorial model content is not available in the backend.
Open a comparison with the three ranked models immediately above and below this model.
Recommendations prioritize the same model type and family, then the closest published LLMBoard score.
Common questions about Kimi K2.5.
Kimi K2.5's default version was released on Jan 27, 2026.
Kimi K2.5's official API price is $0.60 per million input tokens and $3 per million output tokens via Moonshot AI. The lowest tracked third-party offer starts at $0.35 input and $1.7 output via CrofAI.
Kimi K2.5 is published under Moonshot AI in the model registry.
The default version has a 262.1K token context window.
Yes. The default version is marked as open weight under MIT.
23 published provider offerings are linked to the default version.
Nearby ranked alternatives include Qwen3.6 27B, Nemotron 3 Ultra, DeepSeek-V4-Flash.