Reasoning
198-question Diamond subset; independent AA evaluation
- 01
- 02
- 03
- 03
- 03
- 26
Kimi K2.5
Across the task benchmarks we track, Kimi K2.5’s strongest relative results are in coding, web research, and terminal use. On SWE-bench Verified, it scores 76.8%, ranking #5 among 7 models with reported results in our dataset.
Updated: Data as of:
Strongest results
| Benchmark | Measures | Score |
|---|---|---|
| Humanity's Last Exam ↗Stable HLE — text-only subset | Expert reasoning | 30.72%#47 of 68 |
| BrowseComp ↗Published open-model reports | Web research | 74.9%#6 of 8 |
| Terminal-Bench ↗2.1 / Artificial Analysis | Working in a terminal | 45.69%#10 of 13 |
| IFBench ↗58 constraint types | Precise instruction following | 70.2%#13 of 17 |
Model details
| Model | Kimi K2.5 |
|---|---|
| Agent Index v1.2 | Insufficient coverage (0/7) · Preview |
| Creator | Moonshot AI (opens in a new tab) |
| Availability | Open source / open weights |
| Context size | 256,000 tokens (opens in a new tab) |
| Max output | Not reported |
| Knowledge cutoff | Not reported |
| License | Modified MIT License · Model weights ↗ (opens in a new tab) |
Pricing and performance
Per provider and reported configuration. Prices are USD per 1M tokens.
No provider pricing or performance reports have been added for this model yet.
Benchmark performance
Kimi K2.5 in color, alongside leading models. Latest stored reports, best effort per report date.
GPQA Diamond ↗
Coding
Verified split; published reports
- 01
- 02
- 03
- 04
- 05
Terminal Use
89 tasks; Terminus 2; e2b; three repeats
- 01
- 02
- 03
- 04
- 05
- 10
Web Research
Published reports; source settings retained
- 01
- 02
- 03
- 04
- 05
- 06
Knowledge Work
220 tasks; Stirrup agent; pairwise Elo
- 01
- 02
- 03
- 04
- 05
- 34
Reported benchmarks
All published results in our dataset, with versions, evaluation settings, and sources.
| Benchmark | Score | Evaluation setup | Source |
|---|---|---|---|
| BrowseCompPublished open-model reportsAccuracy · Higher is better | 74.9% | harness comparison: provider-reported; settings differ across models · Single-agent context-managed track; search, code and browsing tools; excludes 78.4 agent-swarm score | Moonshot AI ↗ (opens in a new tab)provider reported |
| Gaia2CLI · OpenClaw 2026.4.1Pass@1 · Higher is better | 34% | api host: OpenRouter · capabilities: search, execution, adaptability, ambiguity, time · edition: Gaia2-CLI · harness: OpenClaw 2026.4.1 · reasoning: enabled · reasoning context: Not round-tripped by this harness for OpenRouter; may affect multi-step performance | Meta Agents Research Environments ↗ (opens in a new tab)benchmark publisher · 2026-04-14 |
| GDPval-AA2.1Elo · Higher is better | 846.64 Elo | Kimi K2.5 (Reasoning); 220 tasks; Stirrup agent; pairwise Elo | Artificial Analysis (opens in a new tab)independent |
| GPQADiamond / Artificial AnalysisAccuracy · Higher is better | 78.889% | Kimi K2.5 (Non-reasoning); 198-question Diamond subset; independent AA evaluation | Artificial Analysis (opens in a new tab)independent |
| GPQADiamond / Artificial AnalysisAccuracy · Higher is better | 87.879% | Kimi K2.5 (Reasoning); 198-question Diamond subset; independent AA evaluation | Artificial Analysis (opens in a new tab)independent |
| Humanity's Last ExamStable HLE — full setAccuracy · Higher is better | 30.1% | Full text and image set; no tools; 96K completion · No tools | Moonshot AI ↗ (opens in a new tab)provider reported |
| Humanity's Last ExamStable HLE — full setAccuracy · Higher is better | 50.2% | Full text and image set; search, code and browsing; 96K completion; simple context management; Hugging Face blocked · Tools enabled | Moonshot AI ↗ (opens in a new tab)provider reported |
| Humanity's Last ExamStable HLE — text-only subsetAccuracy · Higher is better | 13.16% | Kimi K2.5 (Non-reasoning); original HLE text-only 2,158 questions; no tools; not HLE Diamond · No tools | Artificial Analysis (opens in a new tab)independent |
| Humanity's Last ExamStable HLE — text-only subsetAccuracy · Higher is better | 30.723% | Kimi K2.5 (Reasoning); original HLE text-only 2,158 questions; no tools; not HLE Diamond · No tools | Artificial Analysis (opens in a new tab)independent |
| IFBench58 constraint typesAccuracy · Higher is better | 70.204% | selection: best reported configuration per release · Kimi K2.5 (Reasoning) · tools: benchmark-defined | Artificial Analysis (opens in a new tab)independent |
| SWE-benchVerified / mini-SWE-agent 2.0.0 / Bash OnlyResolved · Higher is better | 70.8% | attempts: 1 · harness: mini-SWE-agent · harness version: 2.0.0 · reasoning effort: high · Official Bash Only leaderboard; mini-SWE-agent 2.0.0; 500 tasks; one attempt; effort high · tasks: 500 | SWE-bench ↗ (opens in a new tab)benchmark publisher · 2026-02-17 |
| SWE-benchVerified / published open-model reportsResolved · Higher is better | 76.8% | harness comparison: provider-reported; settings differ across models · Internal minimal coding agent; non-thinking; average of five independent runs | Moonshot AI ↗ (opens in a new tab)provider reported |
| Terminal-Bench2.0 / published open-model reportsTask success · Higher is better | 50.8% | harness comparison: provider-reported; settings differ across models · Published Terminal-Bench 2.0 score; not 2.1 | Moonshot AI ↗ (opens in a new tab)provider reported |
| Terminal-Bench2.1 / Artificial AnalysisPass@1 · Higher is better | 45.693% | Kimi K2.5 (Reasoning); 89 tasks; Terminus 2; e2b; three repeats | Artificial Analysis (opens in a new tab)independent |
| π-BenchOverall · PROCProactivity (PROC) · Higher is better | 61.4% | completeness score: 53.9 · completeness uncertainty: 0.8 · metric: Avg PROC · personas: 5 · proc uncertainty: 2.1 · selection: Overall rows; not per-persona averages. COMP is a separate metric and does not determine this ranking. · tasks: 100 | π-Bench research team ↗ (opens in a new tab)benchmark publisher |