Moonshot AIKimi K2.5
Kimi K2.5 is a Moonshot AI open-weights reasoning model, released 27 Jan 2026. Its best configuration (thinking reasoning effort) ranks #172 of 785 on the BenchLeader Index at 57.6 ±3.6, in the upper half. The ± is the point: 242 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 instruction following (69) and lowest in agents & tools (45). At $1.20 per million tokens blended it is pricier than most ranked models. Output speed of 68 tokens per second puts it slower than most, with a first answer in 45.3 s. It has been measured at 5 reasoning-effort settings; this summary describes the best-scoring one, and the tabs above switch between them. Last measured 23 Oct 2026.
- Blended price
- $1.20/M
- $0.600 in · $3.00 out
- Output speed
- 68 tok/s
- measured by Artificial Analysis
- First answer
- 45 s
- first token 1.31 s
- Context
- 262k
- Full answer
- 53 s
- median, reasoning included
- Cost per run
- –
- Released
- 27 Jan 2026
- Overall index58
- Reasoning57
- Coding56
- Agents & tools45
- Maths56
- Knowledge60
- Instruction following69
- Human preference64
- Multimodal60
- Long context64
- Composite56
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.
Reasoning-effort configurations
The same model behaves differently depending on how much it is allowed to think. Each row is one setting, scored only on the benchmarks that were run at that setting. “Not stated” collects results from publishers that did not say which setting they used; for a reasoning model that is usually its thinking mode, but we do not assume it. Pick a setting here or at the top of the page to see its price, speed and category scores.
| Effort | Index | Rank | Speed | First answer | Chat reply cost | Categories |
|---|---|---|---|---|---|---|
| no reasoning | 50.9 | #354 | 52 tok/s | 2.38 s | $0.0011 | Agents & tools 50 · Composite 51 · Instruction following 46 · Knowledge 55 · Long context 59 · Multimodal 56 · Reasoning 42 |
| high | – | – | 36 tok/s | 1.31 s | $0.0011 | Coding 62 |
| max | – | – | 36 tok/s | 1.31 s | $0.0011 | |
| thinkingbest | 57.6 | #172 | 68 tok/s | 45 s | $0.0011 | Agents & tools 45 · Coding 56 · Composite 56 · Human preference 64 · Instruction following 69 · Knowledge 60 · Long context 64 · Maths 56 · Multimodal 60 · Reasoning 57 |
| not stated | 54.5 | #252 | 36 tok/s | 1.31 s | $0.0011 | Agents & tools 46 · Coding 51 · Human preference 61 · Instruction following 55 · Knowledge 48 · Long context 65 · Maths 63 · Multimodal 54 · Reasoning 53 |
How its index has moved
8 Sept 2026 to 21 Oct 2026- thinking
- not stated
- no reasoning
The 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 | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| GPQA Diamondeffort not stated | – | – | 87.6%#72 | Epoch AI Benchmarking Hub | |
| Humanity's Last Exameffort not stated | – | – | 24.4%#18 | Scale AI / CAIS | |
| SimpleBencheffort not stated | – | – | 46.8%#48 | SimpleBench | |
| Arena Hard Prompts | – | 1471#81 | 1460#95 | Arena (LMArena) | |
| GPQA Diamond (AA)not in index | 78.9%#206 | 87.9%#94 | – | Artificial Analysis | |
| Humanity's Last Exam (AA)not in index | 13.2%#273 | 30.7%#145 | – | Artificial Analysis | |
| GPQA Diamond (Vals)not in index | – | 84.1%#52 | – | Vals AI | |
| Kagi LLM Benchmark | – | 78.5%#10 | 63.8%#40 | Kagi LLM Benchmark | |
| ARC-AGI-1effort not stated | – | – | 65.3%#154 | ARC Prize | |
| ARC-AGI-2effort not stated | – | – | 11.8%#158 | ARC Prize | |
| CritPt | 0.6%#254 | 3.1%#176 | – | Artificial Analysis | |
| CL-benchnot in indexeffort not stated | – | – | 19.3%#10 | Epoch AI Benchmarking Hub | |
| CL-bench Lifenot in indexeffort not stated | – | – | 13.2%#10 | Epoch AI Benchmarking Hub |
Coding
| Benchmark | high | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SWE-bench Verified (Epoch)effort not stated | – | – | 73.8%#17 | Epoch AI Benchmarking Hub | |
| SciCodeeffort not stated | – | – | 49.0%#118 | SciCode | |
| WeirdMLeffort not stated | – | – | 45.6%#84 | WeirdML | |
| Arena Coding | – | 1502#74 | 1504#70 | Arena (LMArena) | |
| Arena WebDev | – | 1437#78 | 1406#88 | Arena (LMArena) | |
| LiveCodeBench | – | 83.9%#44 | – | Vals AI | |
| SWE-bench (Vals)not in index | – | 70.0%#58 | – | Vals AI | |
| SWE-bench Verified (bash only) | 70.8%#11 | – | – | SWE-bench | |
| SWE-bench Verified (any scaffold)not in index | 70.8%#14 | – | – | SWE-bench | |
| Code Migrationnot in index | – | 7.0%#64 | – | Vals AI | |
| Vibe Code Bench v1.1not in index | – | 17.5%#88 | – | Vals AI | |
| SWE Atlas: Codebase QnAnot in indexeffort not stated | – | – | 13.1%#20 | Scale AI SEAL | |
| SWE Atlas: Refactoringnot in indexeffort not stated | – | – | 20.9%#15 | Scale AI SEAL | |
| SWE Atlas: Test Writingnot in indexeffort not stated | – | – | 25.8%#21 | Scale AI SEAL | |
| ALE-Benchnot in indexeffort not stated | – | – | 821.6#70 | Epoch AI Benchmarking Hub | |
| Terminal-Bench 2.1 (AA)not in index | – | 45.7%#126 | – | Artificial Analysis |
Agents & tools
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Terminal-Bencheffort not stated | – | – | 43.2%#45 | Terminal-Bench | |
| Terminal-Bench Hard | 18.9%#166 | 34.9%#82 | – | Artificial Analysis | |
| τ²-Bench Telecom (AA)not in index | 81.3%#118 | 95.9%#15 | – | Artificial Analysis | |
| Terminal-Bench 2.1 (Vals) | – | 42.0%#64 | – | Vals AI | |
| MCP Atlaseffort not stated | – | – | 64.4%#24 | Scale AI SEAL | |
| τ³-Banking (AA) | – | 14.2%#123 | – | Artificial Analysis | |
| APEX-Agents (AA)not in index | – | 11.5%#28 | – | Artificial Analysis | |
| Finance Agent v2not in index | – | 35.8%#66 | – | Vals AI | |
| Harvey's Legal Agent Benchmark v1not in index | – | 0.0%#61 | – | Vals AI | |
| Legal Research Benchnot in index | – | 15.9%#62 | – | Vals AI | |
| Terminal-Bench 2.0 (Vals)not in index | – | 40.5%#32 | – | Vals AI | |
| Vending-Bench 2not in indexeffort not stated | – | – | 1198.5#53 | Andon Labs |
Maths
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| OTIS Mock AIMEeffort not stated | – | 92.2%#67 | Epoch AI Benchmarking Hub | |
| AIME (Vals) | 95.6%#8 | – | Vals AI | |
| AIME 2026 | 95.8%#13 | – | MathArena | |
| HMMT February 2026 | 87.1%#19 | – | MathArena | |
| MathArena Apexpossible contamination | 8.8%#23 | – | MathArena | |
| Arena Maths | 1471#51 | 1442#92 | Arena (LMArena) |
Knowledge
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| SimpleQA Verifiedeffort not stated | – | – | 34.3%#53 | Epoch AI Benchmarking Hub | |
| AA-Omniscience | -13.8#216 | -7.3#174 | – | Artificial Analysis | |
| MMLU-Pro | – | 85.9%#50 | – | Vals AI | |
| CorpFin | – | 68.3%#9 | – | Vals AI | |
| TaxEval | – | 74.2%#39 | – | Vals AI | |
| MedQA | – | 94.4%#17 | – | Vals AI | |
| PRBench Financeeffort not stated | – | – | 46.5%#17 | Scale AI SEAL | |
| PRBench Legaleffort not stated | – | – | 43.8%#25 | Scale AI SEAL | |
| MultiNRCeffort not statedpossible contamination | – | – | 35.2%#21 | Scale AI SEAL | |
| CaseLaw v2not in index | – | 58.7%#28 | – | Vals AI | |
| Excel Modeling Benchmarknot in index | – | 28.5%#63 | – | Vals AI | |
| MedCodenot in index | – | 39.3%#66 | – | Vals AI | |
| MedScribenot in index | – | 76.4%#74 | – | Vals AI | |
| AA-Omniscience: accuracynot in index | 24.1%#250 | 35.2%#152 | – | Artificial Analysis | |
| AA-Omniscience: non-hallucinationnot in index | 50.0%#115 | 34.3%#193 | – | Artificial Analysis |
Instruction following
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| IFBench | 43.7%#220 | 70.2%#69 | – | Artificial Analysis | |
| MultiChallengeeffort not stated | – | – | 61.4%#13 | Scale AI SEAL | |
| TutorBencheffort not stated | – | – | 54.6%#6 | Scale AI SEAL | |
| Arena Instruction Followingnot in index | – | 1439#79 | 1432#92 | Arena (LMArena) |
Human preference
| Benchmark | thinking | not stated | Source | Trend |
|---|---|---|---|---|
| Arena Text | 1450#76 | 1431#109 | Arena (LMArena) | |
| Arena Creative Writingnot in index | 1423#78 | 1393#115 | Arena (LMArena) | |
| Arena Multi-turnnot in index | 1452#84 | 1440#103 | Arena (LMArena) | |
| Arena Longer Queriesnot in index | 1458#75 | 1443#98 | Arena (LMArena) |
Multimodal
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Arena Vision | – | 1269#46 | 1254#60 | Arena (LMArena) | |
| MMMU-Pro | 73.1%#127 | 75.4%#102 | – | Artificial Analysis | |
| VISTAeffort not stated | – | – | 41.9%#37 | Scale AI SEAL | |
| MMMU-Pro (Vals)not in index | – | 84.3%#26 | – | Vals AI | |
| MortgageTaxnot in index | – | 66.5%#33 | – | Vals AI | |
| SAGEnot in index | – | 49.9%#19 | – | Vals AI |
Long context
| Benchmark | no reasoning | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|
| Fiction.LiveBench 120keffort not stated | – | – | 78.1%#6 | Fiction.live | |
| AA-LCR | 67.3%#240 | 78.0%#121 | – | Artificial Analysis | |
| Arena Documentnot in index | – | 1436#37 | – | Arena (LMArena) |
Composite
| Benchmark | no reasoning | max | thinking | not stated | Source | Trend |
|---|---|---|---|---|---|---|
| Epoch Capabilities Indexnot in indexeffort not stated | – | – | – | 148.1#66 | Epoch AI Benchmarking Hub | |
| AA Intelligence Index v4.3.2 | 19.4#241 | – | 23.5#187 | – | Artificial Analysis | |
| Vals RSI Indexnot in index | – | 6.8%#25 | – | – | Vals AI |
Safety & honesty
| Benchmark | not stated | Source | Trend |
|---|---|---|---|
| FORTRESSnot in index | 41.1%#19 | Scale AI SEAL | |
| MASKnot in indexpossible contamination | 70.5%#31 | Scale AI SEAL |
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 |
|---|---|---|---|---|---|---|
| Amazon Bedrock | 99 tok/s | 0.90 s | $0.600 | $3.00 | 262k | – |
| Venice | 52 tok/s | 0.97 s | $0.532 | $3.33 | 256k | – |
| AtlasCloud | 45 tok/s | 0.80 s | $0.490 | $2.50 | 262k | int4 |
| NovitaAI | 42 tok/s | 2.46 s | $0.570 | $2.85 | 262k | – |
| SiliconFlow | 25 tok/s | 1.92 s | $0.500 | $2.50 | 262k | int4 |
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 at $0.100 per 1M. Reasoning tokens are not modelled.
| Workload | Tokens in / out | Cost | With caching | Time |
|---|---|---|---|---|
| Chat reply | 400 / 300 | $0.0011 | $0.0010 | 49.7 s |
| Summarize a 30-page report | 12,000 / 600 | $0.0090 | $0.0045 | 54.0 s |
| Code edit | 6,000 / 1,500 | $0.0081 | $0.0059 | 1.1 min |
| Agentic coding session | 60,000 / 4,000 | $0.048 | $0.025 | 1.7 min |
| Structured extraction | 2,000 / 200 | $0.0018 | $0.0011 | 48.2 s |
More from Moonshot AI
See also
Questions people ask
- What are Kimi K2.5's benchmark scores?
- As of 25 Oct 2026, Kimi K2.5 (thinking effort) scores 57.6 ±3.6 on the BenchLeader Index, strongest in instruction following (69) and weakest in agents & tools (45). That rests on 25 independent benchmark results, and every score on this page links to its publisher.
- How does Kimi K2.5 rank?
- Kimi K2.5 (thinking effort) is #172 of 785 ranked configurations on 25 Oct 2026. 242 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 K2.5 cost per 1M tokens?
- As of 25 Oct 2026, Kimi K2.5 costs $0.600 per 1M input tokens and $3.00 per 1M output tokens, $1.20 blended (three input tokens per output token). The list price last changed on 21 Oct 2026.
- When was Kimi K2.5 released?
- Moonshot AI released Kimi K2.5 on 27 Jan 2026. BenchLeader has results for it at 5 reasoning-effort settings.
Data as of 25 Oct 2026. Compare these configurations.
Cite as: BenchLeader, “Kimi K2.5: benchmarks, pricing, speed and rank”, data as of 25 Oct 2026, https://www.benchleader.com/models/kimi-k2-5.