Kimi K2.7 Code leads on the shared-benchmark average
Ahead on 3 of 3 shared benchmarks, averaging 8.5 points higher
Summarised only from the 3 percentage-scale benchmarks scored by every selected model; details are below.
“Best available” takes each model’s highest recorded non-parallel mode per benchmark, so it may combine modes into a virtual configuration that does not exist. Read it with the mode breakdown.
Kimi K2.5
Moonshot AI
Benchmark-by-benchmark comparison. Changing the thinking mode or tool filters updates the chart and table below.
“Best available” picks the highest non-parallel mode separately for each benchmark. The resulting series can combine several reasoning levels and is not one reproducible runtime configuration. Choose a mode filter to compare like-for-like runs.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
Each axis is the mean percentage score of one benchmark domain. It is an average, not a capability rating.
Relative edge: none clear / Relative gap: AI Agent - 工具使用 -11.6
Relative edge: AI Agent - 工具使用 +11.6 / Relative gap: none clear
Method: for each model and benchmark, all scores in the current mode scope are averaged (not the best score), then those benchmark scores are averaged within each domain. Only benchmarks scored on a 0-100 scale by at least two of the selected models count — Elo and rating-scale benchmarks such as Codeforces or Arena are excluded, because averaging a 1500 rating with an 85% accuracy produces a meaningless number. Missing values are not counted as zero, and the averages are unweighted, so domains with harder benchmarks read lower.
Every model and runtime mode, benchmark by benchmark. Values are comparable along a row, not between different benchmarks.
3 benchmarks with comparable scores. Each model shows its best score; mode label is displayed below.
| Benchmark | Kimi K2.5 | Kimi K2.7 Code |
|---|---|---|
LiveBench 综合评估 | 69.07Thinking Enabled | 71.89Standard Mode |
SimpleBench 常识推理 | 46.80Thinking Enabled | 57.90Thinking Enabled |
MCP-Atlas AI Agent - 工具使用 | 64.40Standard Mode ｜ Tools | 76.00Thinking Enabled ｜ Tools |
Official list prices per model API, split by input and output. Unit: USD per 1M tokens.
Architecture, licensing and API modalities. "Not provided" means the field is missing from our database.
| Features & specs | Kimi K2.5Moonshot AI | Kimi K2.7 CodeMoonshot AI |
|---|---|---|
Core specsRelease | 2026-01-27 | 2026-06-12 |
Context length | 256K | 256K |
Total parameters | 1T | 1T |
Active parameters | 32B | 32B |
Max output length | 16,384 tokens | Not provided |
Architecture | MoE (mixture of experts) | MoE (mixture of experts) |
Runtime modes | 关闭开启扩展 | 开启 |
Availability & licensingCode availability | Available · Modified MIT License | Available · Modified MIT |
Weight availability | Available · Modified MIT License | Available · Modified MIT |
Use & commercial terms | 免费商用授权 | 免费商用授权 |
Local deploymentWeight size | 595GB | 595GB |
VRAM for weights | ≈ 595 GB (weights only, excludes KV cache) | ≈ 595 GB (weights only, excludes KV cache) |
Weights | Hugging Face | Hugging Face |
Source repo | GitHub | GitHub |
API modality supportText Input/Output | Input：YesOutput：Yes | Input：YesOutput：Yes |
Image Input/Output | Input：YesOutput：No | Input：YesOutput：No |
Video Input/Output | Input：NoOutput：No | Input：YesOutput：No |
ResourcesPaper / report | Kimi K2.5: Visual Agentic Intelligence | Kimi K2.7 Code |
DataLearner blog | 重磅！Kimi K2.5发布，依然免费开源！原生多模态MoE架构，全球最大规模参数的开源模型之一，官方评测结果比肩诸多闭源模型！可以驱动100个子Agent执行！ | Not provided |
Kimi K2.7 Code
Moonshot AI
