Moonshot AIKimi K2
Kimi K2 is a Moonshot AI open-weights model, released 11 Jul 2025. Its best configuration ranks #371 of 785 on the BenchLeader Index at 50.4 ±3.5, in the upper half. The ± is the point: 302 other configurations score within that range, so they and this one cannot be told apart on quality alone — price and speed are what separate them. It scores highest in maths (61) and lowest in reasoning (41). At $1.00 per million tokens blended it is mid-priced. Output speed of 47 tokens per second puts it in the slowest quarter, with a first answer in 1.9 s. It has been superseded by Kimi K2 0905, which is the version to compare against today. Last measured 22 Oct 2026.
- Blended price
- $1.00/M
- $0.570 in · $2.30 out
- Output speed
- 47 tok/s
- measured by Artificial Analysis
- First answer
- 1.94 s
- first token 1.34 s
- Context
- 131k
- Full answer
- 12 s
- median, reasoning included
- Cost per run
- –
- Released
- 11 Jul 2025
- Overall index50
- Reasoning41
- Coding48
- Agents & tools52
- Maths61
- Knowledge45
- Instruction following53
- Human preference60
- Long context51
- Composite44
Versions
Moonshot AI has shipped 2 models under this name. Each is ranked on its own results; a newer version often has fewer results so far, which holds its index nearer the average until more arrive.
| Model | Released | Index | Rank |
|---|---|---|---|
| Kimi K2 0905 | 5 Sept 2025 | 50.5 | #367 |
| Kimi K2this page | 11 Jul 2025 | 50.4 | #371 |
How its index has moved
8 Sept 2026 to 21 Oct 2026The index is recomputed from scratch every day, so a line moves when a new benchmark result lands, when a publisher revises a score, or when the models it is normalised against change. Early movement usually means the score is still settling.
Benchmark results
See all LLM benchmarks. One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.
Reasoning
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| SimpleBench | 26.3%#73 | SimpleBench | |
| Arena Hard Prompts | 1432#135 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 76.6%#232 | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 7.4%#365 | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | 71.5%#87 | Vals AI | |
| Kagi LLM Benchmark | 45.0%#92 | Kagi LLM Benchmark | |
| GPQA Diamond (HELM)not in index | 65.2%#19 | HELM Capabilities | |
| CritPt | 0.0%#315 | Artificial Analysis | |
| ForecastBenchnot in index | 62.2%#59 | ForecastBench |
Coding
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| WeirdML | 39.4%#108 | WeirdML | |
| GSO-Bench | 4.9%#25 | GSO-Bench | |
| Arena Coding | 1460#131 | Arena (LMArena) | |
| LiveCodeBench | 70.5%#85 | Vals AI | |
| SWE-Bench Pro | 27.7%#17 | Scale AI SEAL | |
| Aider Polyglot | 59.1%#17 | Aider polyglot leaderboard | |
| SWE-bench Verified (bash only) | 43.8%#35 | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 65.4%#21 | SWE-bench | |
| SWE-Bench Pro (private)not in index | 27.7%#17 | Scale AI SEAL |
Agents & tools
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Terminal-Bench | 27.8%#61 | Terminal-Bench | |
| Terminal-Bench Hard | 15.9%#190 | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 61.1%#177 | Artificial Analysis | |
| BFCL Overall | 59.1%#9 | Berkeley Function Calling Leaderboard | |
| Terminal-Bench 2.0 (Vals)not in index | 25.8%#50 | Vals AI | |
| BrowseComp-Plusnot in index | 14.6%#15 | BrowseComp-Plus |
Maths
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AIME (Vals) | 62.7%#55 | Vals AI | |
| MGSM | 91.0%#29 | Vals AI | |
| Omni-MATH (HELM) | 65.4%#6 | HELM Capabilities | |
| MATH 500not in index | 94.2%#10 | Vals AI | |
| Arena Maths | 1388#165 | Arena (LMArena) |
Knowledge
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-Omniscience | -28.3#273 | Artificial Analysis | |
| MMLU-Pro | 79.4%#92 | Vals AI | |
| LegalBench | 81.5%#80 | Vals AI | |
| CorpFin | 50.4%#102 | Vals AI | |
| TaxEval | 70.2%#89 | Vals AI | |
| MedQA | 84.0%#59 | Vals AI | |
| PRBench Finance | 38.3%#32 | Scale AI SEAL | |
| PRBench Legal | 36.4%#39 | Scale AI SEAL | |
| MultiNRC | 18.5%#34 | Scale AI SEAL | |
| MMLU-Pro (HELM)not in index | 81.9%#14 | HELM Capabilities | |
| AA-Omniscience: accuracynot in index | 27.4%#206 | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 23.4%#255 | Artificial Analysis |
Instruction following
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| IFBench | 41.5%#246 | Artificial Analysis | |
| IFEval (HELM) | 85.0%#19 | HELM Capabilities | |
| WildBench (HELM) | 86.2%#3 | HELM Capabilities | |
| Arena Instruction Followingnot in index | 1381#162 | Arena (LMArena) |
Human preference
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Arena Text | 1418#127 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1371#147 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1420#131 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1395#165 | Arena (LMArena) | |
| Lech Mazur Creative Writingnot in index | 85.6%#2 | Lech Mazur |
Long context
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| AA-LCR | 53.0%#309 | Artificial Analysis |
Composite
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| Epoch Capabilities Indexnot in index | 140.0#112 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 12.7#324 | Artificial Analysis | |
| HELM Capabilities meannot in index | 76.8%#10 | HELM Capabilities |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 55.5%#9 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 46.7%#60 | Scale AI SEAL |
Where it wins
Benchmarks where this configuration ranks in the top five of every configuration measured.
Where to run it
Every provider serving this model through OpenRouter, with throughput and first-token latency measured on live traffic over the last 30 minutes and each provider’s own price. Purple marks the best in each column.
| Provider | Speed | First token | Input $/M | Output $/M | Context | Quantisation |
|---|---|---|---|---|---|---|
| NovitaAI | 41 tok/s | 1.21 s | $0.570 | $2.30 | 131k | fp8 |
Price history
The price above, per 1M tokens, every day it changed.
What a task costs
Estimates from list price, output speed and time to first answer for the best configuration. “With caching” assumes three-quarters of the input is served from the prompt cache. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0009 | – | 8.3 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0082 | – | 14.6 s |
| Code edit | 6,000 / 1,500 | $0.0069 | – | 33.6 s |
| Agentic coding session | 60,000 / 4,000 | $0.043 | – | 1.4 min |
| Structured extraction | 2,000 / 200 | $0.0016 | – | 6.2 s |
More from Moonshot AI
See also
Questions people ask
- What are Kimi K2's benchmark scores?
- As of 25 Oct 2026, Kimi K2 scores 50.4 ±3.5 on the BenchLeader Index, strongest in maths (61) and weakest in reasoning (41). That rests on 33 independent benchmark results, and every score on this page links to its publisher.
- How does Kimi K2 rank?
- Kimi K2 is #371 of 785 ranked configurations on 25 Oct 2026. 302 other configurations score within its margin, so the exact position matters less than the band. The leader is Claude Opus 5.5 (max effort) at 72.6.
- How much does Kimi K2 cost per 1M tokens?
- As of 25 Oct 2026, Kimi K2 costs $0.570 per 1M input tokens and $2.30 per 1M output tokens, $1.00 blended (three input tokens per output token). The list price last changed on 9 Sept 2026.
- When was Kimi K2 released?
- Moonshot AI released Kimi K2 on 11 Jul 2025.
Data as of 25 Oct 2026. Compare with another model.
Cite as: BenchLeader, “Kimi K2: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/kimi-k2.