Kimi K2.5 vs Kimi K3.1 Pro
Kimi K3.1 Pro leads Kimi K2.5 on coding, 82.2 to 38.6 on the Vibe Coding Index, and Kimi K2.5 costs 3.5x less per output token ($2.85 vs $10.00 per million). Updated October 29, 2026; 16 benchmarks have a result for both.
Kimi K2.5 or Kimi K3.1 Pro: which to pick
Pick Kimi K3.1 Pro for coding: it scores 82.2 on the Vibe Coding Index against 38.6 for Kimi K2.5.
Pick Kimi K2.5 for output tokens 3.5x cheaper ($2.85 vs $10.00 per million).
- Shared results
- 16
- Kimi K2.5 only
- 89
- Kimi K3.1 Pro only
- 19
- Like-for-like categories
- 6 of 9
Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.
Which one for your work
Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.
Coding
Vibe Coding Index, six public coding benchmarks
Kimi K3.1 Pro
Kimi K3.1 Pro scores 82.2 against 38.6.
3 and 4 benchmarks behind the scores
Intelligence
AA Intelligence Index
No call
Only Kimi K3.1 Pro has a AA Intelligence Index result, 46.0.
Directional only
Reasoning
GPQA Diamond
No call
Only Kimi K2.5 has a GPQA Diamond result, 84.1%.
Directional only
Math
AIME
No call
Only Kimi K2.5 has a AIME result, 95.6%.
Directional only
Knowledge
MMLU Pro
No call
Only Kimi K2.5 has a MMLU Pro result, 85.9%.
Directional only
Web design
Design Arena: websites
Kimi K3.1 Pro
Kimi K3.1 Pro leads, 1325 against 1247.
Same board, both measured
General
Vals Index
No call
Only Kimi K3.1 Pro has a Vals Index result, 62.5%.
Directional only
Long documents
Listed context window
Kimi K3.1 Pro
Kimi K3.1 Pro takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Kimi K2.5
Kimi K2.5 $0.0020, Kimi K3.1 Pro $0.0070 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Kimi K2.5 has the number, $0.24.
Directional only
At a glance
- Coding index
- Kimi K2.5 38.6 from 3 benchmarks, Kimi K3.1 Pro 82.2 from 4 benchmarks.
- Widest category
- Math, 42.0 points apart on average over 1 shared benchmark.
- Widest single gap
- Vibe Code Bench (Vals AI): Kimi K2.5 17.5% against Kimi K3.1 Pro 84.9%, 67.3 points apart on a 0 to 100 scale.
- Output price
- Kimi K2.5 $2.85 against Kimi K3.1 Pro $10.00 per million output tokens, 3.5x apart.
- Context window
- Kimi K2.5 takes 262K tokens, Kimi K3.1 Pro 1.05M.
Benchmark by benchmark
First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.
| Category | Kimi K2.5 | Bars | Kimi K3.1 Pro | Ties |
|---|---|---|---|---|
| Coding | 0 | 5 | ||
| Agentic work | 0 | 1 | ||
| Reasoning | 0 | 1 | ||
| Math | 0 | 1 | ||
| Web design | 0 | 7 | ||
| General | 0 | 1 | ||
| All shared boards | 0 | 16 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index05
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| SWE-bench Verified | 70.8% | Not measured | Only Kimi K2.5 measured |
| SWE-bench Multilingual | 67.3% | Not measured | Only Kimi K2.5 measured |
| LiveBench coding | Not measured | 81.8% | Only Kimi K3.1 Pro measured |
| LMArena WebDev | 1436 | 1695 | Kimi K3.1 Pro |
| AA Coding Index | 46.8 | Not measured | Only Kimi K2.5 measured |
| LiveCodeBench (Vals AI) | 83.9% | Not measured | Only Kimi K2.5 measured |
| SWE-bench Verified (Vals AI) | 70.0% | Not measured | Only Kimi K2.5 measured |
| Terminal-Bench 4 (Vals AI) | Not measured | 63.6% | Only Kimi K3.1 Pro measured |
| Vibe Code Bench (Vals AI) | 17.5% | 84.9% | Kimi K3.1 Pro |
| Code migration (Vals AI) | 7.0% | Not measured | Only Kimi K2.5 measured |
| IOI (Vals AI) | Not measured | 81.9% | Only Kimi K3.1 Pro measured |
| Terminal-Bench 2.1 (Vals AI) | 42.0% | Not measured | Only Kimi K2.5 measured |
| SciCode | Not measured | 60.8% | Only Kimi K3.1 Pro measured |
| Terminal-Bench Hard (Artificial Analysis run) | 34.8% | Not measured | Only Kimi K2.5 measured |
| Terminal-Bench | 43.2% | Not measured | Only Kimi K2.5 measured |
| WeirdML | 45.6% | Not measured | Only Kimi K2.5 measured |
| ALE-Bench | 821.6 | 1932.8 | Kimi K3.1 Pro |
| LMArena Text: coding | 1483 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: software and IT | 1471 | Not measured | Only Kimi K2.5 measured |
| LMArena WebDev: React | 1429 | 1701 | Kimi K3.1 Pro |
| LMArena WebDev: HTML | 1426 | 1706 | Kimi K3.1 Pro |
| LMArena WebDev: from an image | 1439 | Not measured | Only Kimi K2.5 measured |
| SWE Atlas: Codebase QnA | 13.1% | Not measured | Only Kimi K2.5 measured |
| SWE Atlas: Refactoring | 20.9% | Not measured | Only Kimi K2.5 measured |
| SWE Atlas: Test Writing | 25.8% | Not measured | Only Kimi K2.5 measured |
| Neither model has a result on the other 18 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 46.0 | Only Kimi K3.1 Pro measured |
| Epoch Capabilities Index | 148.1 | Not measured | Only Kimi K2.5 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| Design Arena: agents building web apps | 1137 | Not measured | Only Kimi K2.5 measured |
| Design Arena: agents building full-stack apps | 1114 | Not measured | Only Kimi K2.5 measured |
| Design Arena: agents building mobile apps | 1147 | Not measured | Only Kimi K2.5 measured |
| Design Arena: agents building Android apps | 1115 | Not measured | Only Kimi K2.5 measured |
| Design Arena: agents building Godot games | 1219 | Not measured | Only Kimi K2.5 measured |
| LiveBench research | Not measured | 77.9% | Only Kimi K3.1 Pro measured |
| LiveBench browser | Not measured | 75.9% | Only Kimi K3.1 Pro measured |
| LiveBench long agentic | Not measured | 39.1% | Only Kimi K3.1 Pro measured |
| Finance Agent (Vals AI) | 35.8% | Not measured | Only Kimi K2.5 measured |
| Tax Agent Bench (Vals AI) | Not measured | 24.9% | Only Kimi K3.1 Pro measured |
| Legal Research Bench (Vals AI) | 15.9% | 37.5% | Kimi K3.1 Pro |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 11.7% | Only Kimi K3.1 Pro measured |
| Excel Modeling Benchmark (Vals AI) | 28.5% | Not measured | Only Kimi K2.5 measured |
| BioMysteryBench (Vals AI) | Not measured | 75.6% | Only Kimi K3.1 Pro measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 31.4% | Only Kimi K3.1 Pro measured |
| Vals Science Index | Not measured | 44.4% | Only Kimi K3.1 Pro measured |
| τ²-Bench Telecom | 95.9% | Not measured | Only Kimi K2.5 measured |
| OSWorld | 63.3% | Not measured | Only Kimi K2.5 measured |
| Vending-Bench 2 | $1,198 | Not measured | Only Kimi K2.5 measured |
| MCP Atlas | 64.4% | Not measured | Only Kimi K2.5 measured |
| Neither model has a result on the other 22 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| GPQA Diamond | 84.1% | Not measured | Only Kimi K2.5 measured |
| LiveBench reasoning | Not measured | 91.1% | Only Kimi K3.1 Pro measured |
| MysteryMechanism (Vals AI) | Not measured | 26.1% | Only Kimi K3.1 Pro measured |
| MMMU Pro (Vals AI) | 84.3% | Not measured | Only Kimi K2.5 measured |
| GPQA Diamond (Artificial Analysis run) | 87.9% | Not measured | Only Kimi K2.5 measured |
| Humanity's Last Exam (Artificial Analysis run) | 30.7% | 52.5% | Kimi K3.1 Pro |
| ARC-AGI | 65.3% | Not measured | Only Kimi K2.5 measured |
| ARC-AGI-2 | 11.8% | Not measured | Only Kimi K2.5 measured |
| SimpleBench | 46.8% | Not measured | Only Kimi K2.5 measured |
| LMArena Text: hard prompts | 1453 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: expert | 1466 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: hard prompts in English | 1460 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: diagrams | 1282 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: homework | 1297 | Not measured | Only Kimi K2.5 measured |
| Kagi LLM Benchmark | 78.5% | Not measured | Only Kimi K2.5 measured |
| Neither model has a result on the other 8 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| AIME | 95.6% | Not measured | Only Kimi K2.5 measured |
| ProofBench | Not measured | 91.0% | Only Kimi K3.1 Pro measured |
| LMArena Text: math | 1470 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: mathematical work | 1474 | Not measured | Only Kimi K2.5 measured |
| MathArena | 17.5% | 59.6% | Kimi K3.1 Pro |
| Neither model has a result on the other 9 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| MMLU Pro | 85.9% | Not measured | Only Kimi K2.5 measured |
| CaseLaw (Vals AI) | 58.7% | Not measured | Only Kimi K2.5 measured |
| CorpFin (Vals AI) | 68.3% | Not measured | Only Kimi K2.5 measured |
| TaxEval (Vals AI) | 74.2% | Not measured | Only Kimi K2.5 measured |
| MortgageTax (Vals AI) | 66.5% | Not measured | Only Kimi K2.5 measured |
| MedQA (Vals AI) | 94.4% | Not measured | Only Kimi K2.5 measured |
| MedCode (Vals AI) | 39.3% | Not measured | Only Kimi K2.5 measured |
| SimpleQA Verified | 34.3% | Not measured | Only Kimi K2.5 measured |
| LMArena Text: science | 1459 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: medicine and healthcare | 1448 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: legal and government | 1441 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: business and finance | 1427 | Not measured | Only Kimi K2.5 measured |
| PRBench Finance | 46.5% | Not measured | Only Kimi K2.5 measured |
| PRBench Legal | 43.8% | Not measured | Only Kimi K2.5 measured |
| Neither model has a result on the other 6 boards here | |||
Web designLike-for-likePublic boards, shown as published, kept out of the index07
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| Design Arena: websites | 1247 | 1325 | Kimi K3.1 Pro |
| Design Arena: UI components | 1231 | 1400 | Kimi K3.1 Pro |
| Design Arena: game dev | 1205 | 1410 | Kimi K3.1 Pro |
| Design Arena: data viz | 1220 | 1300 | Kimi K3.1 Pro |
| Design Arena: 3D | 1202 | 1451 | Kimi K3.1 Pro |
| Design Arena: all code categories | 1235 | 1354 | Kimi K3.1 Pro |
| Design Arena: SVG | 1147 | 1328 | Kimi K3.1 Pro |
| Design Arena: ASCII art | 1163 | Not measured | Only Kimi K2.5 measured |
| LiveBench design | Not measured | 77.8% | Only Kimi K3.1 Pro measured |
GeneralLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| Vals Index | Not measured | 62.5% | Only Kimi K3.1 Pro measured |
| LiveBench data analysis | Not measured | 79.1% | Only Kimi K3.1 Pro measured |
| MedScribe (Vals AI) | 76.4% | Not measured | Only Kimi K2.5 measured |
| IFBench | 70.2% | Not measured | Only Kimi K2.5 measured |
| Long context reasoning (AA-LCR) | 78.0% | 86.3% | Kimi K3.1 Pro |
| Fiction.LiveBench | 86.1% | Not measured | Only Kimi K2.5 measured |
| CL-bench | 19.3% | Not measured | Only Kimi K2.5 measured |
| CL-bench Life | 13.2% | Not measured | Only Kimi K2.5 measured |
| LMArena Text | 1445 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: instruction following | 1431 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: creative writing | 1422 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision | 1269 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: English | 1453 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: non-English | 1432 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Chinese | 1494 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: German | 1441 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Japanese | 1422 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Korean | 1412 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: French | 1453 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Spanish | 1450 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Russian | 1435 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: Polish | 1445 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: multi-turn | 1444 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: longer queries | 1444 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: writing and language | 1425 | Not measured | Only Kimi K2.5 measured |
| LMArena Text: entertainment and media | 1417 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: English | 1274 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: Chinese | 1328 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: reading text | 1276 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: humor | 1284 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: captioning | 1257 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: recognizing things | 1285 | Not measured | Only Kimi K2.5 measured |
| LMArena Vision: creative writing | 1269 | Not measured | Only Kimi K2.5 measured |
| MultiChallenge | 61.4% | Not measured | Only Kimi K2.5 measured |
| Neither model has a result on the other 14 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Kimi K2.5 | Kimi K3.1 Pro | Leader |
|---|---|---|---|
| Neither model has a result on 5 boards here | |||
What each workload costs
Three fixed token mixes at each model's listed price, so per-token rates become a bill.
Chat turn
1K input + 500 output tokens
Kimi K2.5 costs less
Repository review
50K input + 3K output tokens
Kimi K2.5 costs less
Agent loop
200K input + 10K output tokens
Kimi K2.5 costs less
Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.
Specification differences
The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.
| Specification | MoonshotAI Kimi K2.5 | MoonshotAI Kimi K3.1 Pro |
|---|---|---|
| Maker | Moonshot AI | Moonshot AI |
| Weights | Open weights | No public weights listed |
| Released | January 27, 2026 | October 23, 2026 |
| Context window | 262K | 1.05M |
| Input $/Mtok | $0.532 | $2.00 |
| Output $/Mtok | $2.85 | $10.00 |
| Coding index | 38.6 | 82.2 |
| Independent measurements behind it | 3 | 4 |
| Cost per solved task | $0.24 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Kimi K2.5 better than Kimi K3.1 Pro for coding?
Pick Kimi K3.1 Pro for coding: it scores 82.2 on the Vibe Coding Index against 38.6 for Kimi K2.5. Pick Kimi K2.5 for output tokens 3.5x cheaper ($2.85 vs $10.00 per million).
Which is cheaper, Kimi K2.5 or Kimi K3.1 Pro?
Kimi K2.5 costs $0.532 per million input tokens and $2.85 per million output tokens; Kimi K3.1 Pro costs $2.00 and $10.00. Cost per solved task is known for Kimi K2.5 only.
Which has the bigger context window, Kimi K2.5 or Kimi K3.1 Pro?
Kimi K3.1 Pro, with 1.05M tokens against 262K.
How many benchmarks is the Kimi K2.5 vs Kimi K3.1 Pro comparison based on?
16 benchmarks have a result for both models as of October 29, 2026. Kimi K2.5's coding index rests on 3 independent benchmarks and Kimi K3.1 Pro's on 4. Where only one model has a result, the page says so instead of naming a winner.