Rankings / Moonshot AI
Kimi K2
released 2025-11-06
BenchAtlas Index
as of 2026-07-14
thinking
rank #264
4 families · 3 categories · medium
thinking
rank #268
5 families · 3 categories · medium
base configuration
rank #310
10 families · 5 categories · high
Benchmark evidence
23 results
agentic coding
| τ²-Bench | 61.1 % | independent | Artificial Analysis |
coding
| LiveCodeBench v6 implementation=artificial-analysis | 55.6 % | independent | Artificial Analysis | |
| SciCode | 34.5 % | independent | Artificial Analysis |
external indices
| Intelligence Index v4.1 | 19.4 points | independent | Artificial Analysis | |
| AA Math Index | 57.0 points | independent | Artificial Analysis | |
| ECI | thinking | 146.0 points | independent | Epoch AI Benchmarking Hub |
knowledge science
| GPQA Diamond implementation=vals-ai | thinking | 78.5 % | independent | Vals AI |
| GPQA Diamond implementation=artificial-analysis | 76.6 % | independent | Artificial Analysis | |
| Humanity's Last Exam implementation=artificial-analysis | 7.0 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=artificial-analysis | 82.4 % | independent | Artificial Analysis | |
| MMLU-Pro implementation=vals-ai | thinking | 81.1 % | independent | Vals AI |
long context instruction
| AA-LCR | 51.0 % | independent | Artificial Analysis | |
| IFBench | 41.5 % | independent | Artificial Analysis | |
| MultiChallenge | thinking | 55.4 % 53.5–57.3 | independent | Scale Labs |
professional
| CaseLaw | thinking | 65.7 % | independent | Vals AI |
| CorpFin | thinking | 60.6 % | independent | Vals AI |
| LegalBench | thinking | 80.2 % | independent | Vals AI |
| MedQA | thinking | 92.6 % | independent | Vals AI |
| TaxEval | thinking | 71.7 % | independent | Vals AI |
reasoning math
| AIME implementation=vals-ai | thinking | 85.4 % | independent | Vals AI |
| AIME implementation=artificial-analysis | 69.3 % | independent | Artificial Analysis | |
| AIME year=2025 · implementation=artificial-analysis | 57.0 % | independent | Artificial Analysis | |
| MATH-500 implementation=artificial-analysis | 97.1 % | independent | Artificial Analysis |
Agent + model results
systems, not bare-model scores
| agent + model Artificial Analysis harness + Kimi K2 | Terminal-Bench Hard | 15.9 % | independent | Artificial Analysis |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.