Devstral 2 2512 vs Kimi K3
Kimi K3 leads Devstral 2 2512 on the Artificial Analysis Intelligence Index, 43.6 to 8.6, and Devstral 2 2512 costs 7.5x less per output token ($2.00 vs $15.00 per million). Updated October 22, 2026; 5 benchmarks have a result for both.
Devstral 2 2512 or Kimi K3: which to pick
Only Kimi K3 has a coding score, 62.3 on the Vibe Coding Index; Devstral 2 2512 has none, so the two cannot be ranked on coding.
Pick Devstral 2 2512 for output tokens 7.5x cheaper ($2.00 vs $15.00 per million).
- Shared results
- 5
- Devstral 2 2512 only
- 1
- Kimi K3 only
- 112
- Like-for-like categories
- 3 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
No call
Only Kimi K3 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Kimi K3
Kimi K3 leads, 43.6 against 8.6.
Same board, both measured
Agentic work
AA Agentic Index
Kimi K3
Kimi K3 leads, 50.0 against 2.3.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Kimi K3 has a GPQA Diamond result, 92.9%.
Directional only
Knowledge
MMLU Pro
No call
Only Kimi K3 has a MMLU Pro result, 88.0%.
Directional only
Web design
Design Arena: websites
No call
Only Kimi K3 has a Design Arena: websites result, 1334.
Directional only
General
Vals Index
No call
Only Kimi K3 has a Vals Index result, 53.0%.
Directional only
Long documents
Listed context window
Kimi K3
Kimi K3 takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Devstral 2 2512
Devstral 2 2512 $0.0014, Kimi K3 $0.011 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Kimi K3
Devstral 2 2512 $1.19, Kimi K3 $0.81.
Published attempt cost and pass rate
At a glance
- Coding index
- Devstral 2 2512 not benchmarked, Kimi K3 62.3 from 6 benchmarks.
- Widest category
- Agentic work, 47.7 points apart on average over 1 shared benchmark.
- Widest single gap
- AA Agentic Index: Devstral 2 2512 2.3 against Kimi K3 50.0, 47.7 points apart on a 0 to 100 scale.
- Output price
- Devstral 2 2512 $2.00 against Kimi K3 $15.00 per million output tokens, 7.5x apart.
- Context window
- Devstral 2 2512 takes 262K tokens, Kimi K3 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 | Devstral 2 2512 | Bars | Kimi K3 | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| All shared boards | 0 | 5 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index03
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| SWE-bench Verified | 53.8% | Not measured | Only Devstral 2 2512 measured |
| LiveBench coding | Not measured | 59.3% | Only Kimi K3 measured |
| LMArena WebDev | Not measured | 1655 | Only Kimi K3 measured |
| AA Coding Index | 31.3 | 76.2 | Kimi K3 |
| LiveCodeBench (Vals AI) | 51.8% | 87.2% | Kimi K3 |
| SWE-bench Verified (Vals AI) | 62.8% | 93.4% | Kimi K3 |
| Terminal-Bench 4 (Vals AI) | Not measured | 17.2% | Only Kimi K3 measured |
| Vibe Code Bench (Vals AI) | Not measured | 85.0% | Only Kimi K3 measured |
| Code migration (Vals AI) | Not measured | 16.1% | Only Kimi K3 measured |
| ProgramBench (Vals AI) | Not measured | 2.0% | Only Kimi K3 measured |
| IOI (Vals AI) | Not measured | 48.9% | Only Kimi K3 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 80.9% | Only Kimi K3 measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 44.2% | Only Kimi K3 measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 18.2% | Only Kimi K3 measured |
| CyberBench (Vals AI) | Not measured | 75.2% | Only Kimi K3 measured |
| SciCode | Not measured | 59.5% | Only Kimi K3 measured |
| FrontierCode | Not measured | 44.2% | Only Kimi K3 measured |
| FrontierSWE | Not measured | 25.9% | Only Kimi K3 measured |
| WeirdML | Not measured | 82.6% | Only Kimi K3 measured |
| ALE-Bench | Not measured | 1524.5 | Only Kimi K3 measured |
| LMArena Text: coding | Not measured | 1508 | Only Kimi K3 measured |
| LMArena Text: software and IT | Not measured | 1499 | Only Kimi K3 measured |
| LMArena WebDev: React | Not measured | 1670 | Only Kimi K3 measured |
| LMArena WebDev: HTML | Not measured | 1644 | Only Kimi K3 measured |
| LMArena WebDev: from an image | Not measured | 1579 | Only Kimi K3 measured |
| Neither model has a result on the other 14 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| AA Intelligence Index | 8.6 | 43.6 | Kimi K3 |
| Epoch Capabilities Index | Not measured | 157.9 | Only Kimi K3 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| AA Agentic Index | 2.3 | 50.0 | Kimi K3 |
| Design Arena: agents building web apps | Not measured | 1295 | Only Kimi K3 measured |
| Design Arena: agents building full-stack apps | Not measured | 1315 | Only Kimi K3 measured |
| Design Arena: agents building mobile apps | Not measured | 1269 | Only Kimi K3 measured |
| Design Arena: agents building Android apps | Not measured | 1253 | Only Kimi K3 measured |
| Design Arena: agents building Godot games | Not measured | 1223 | Only Kimi K3 measured |
| Design Arena: agents building games | Not measured | 1252 | Only Kimi K3 measured |
| Design Arena: HTML slides | Not measured | 1206 | Only Kimi K3 measured |
| Design Arena: PowerPoint slides | Not measured | 1247 | Only Kimi K3 measured |
| Design Arena: spreadsheets | Not measured | 1070 | Only Kimi K3 measured |
| LiveBench research | Not measured | 79.0% | Only Kimi K3 measured |
| LiveBench browser | Not measured | 16.0% | Only Kimi K3 measured |
| LiveBench long agentic | Not measured | 42.8% | Only Kimi K3 measured |
| Finance Agent (Vals AI) | Not measured | 53.1% | Only Kimi K3 measured |
| Tax Agent Bench (Vals AI) | Not measured | 33.4% | Only Kimi K3 measured |
| Legal Research Bench (Vals AI) | Not measured | 46.1% | Only Kimi K3 measured |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 12.9% | Only Kimi K3 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 66.7% | Only Kimi K3 measured |
| BioMysteryBench (Vals AI) | Not measured | 72.6% | Only Kimi K3 measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 1.4% | Only Kimi K3 measured |
| Time Horizon Index (Vals AI) | Not measured | 10.5% | Only Kimi K3 measured |
| APEX-Agents | Not measured | 47.8% | Only Kimi K3 measured |
| PostTrainBench | Not measured | 32.0% | Only Kimi K3 measured |
| Vending-Bench 2 | Not measured | $5,165 | Only Kimi K3 measured |
| Neither model has a result on the other 17 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 92.9% | Only Kimi K3 measured |
| LiveBench reasoning | Not measured | 90.7% | Only Kimi K3 measured |
| MMMU Pro (Vals AI) | Not measured | 88.2% | Only Kimi K3 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 93.5% | Only Kimi K3 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 46.9% | Only Kimi K3 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 93.1% | Only Kimi K3 measured |
| ARC-AGI | Not measured | 94.5% | Only Kimi K3 measured |
| ARC-AGI-2 | Not measured | 60.4% | Only Kimi K3 measured |
| SimpleBench | Not measured | 60.7% | Only Kimi K3 measured |
| Chess puzzles | Not measured | 37.0% | Only Kimi K3 measured |
| Mystery game puzzles | Not measured | 26.0% | Only Kimi K3 measured |
| LMArena Text: hard prompts | Not measured | 1495 | Only Kimi K3 measured |
| LMArena Text: expert | Not measured | 1520 | Only Kimi K3 measured |
| LMArena Text: hard prompts in English | Not measured | 1496 | Only Kimi K3 measured |
| Neither model has a result on the other 8 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 97.2% | Only Kimi K3 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 72.2% | Only Kimi K3 measured |
| FrontierMath, tier 4 | Not measured | 39.0% | Only Kimi K3 measured |
| ProofBench | Not measured | 87.0% | Only Kimi K3 measured |
| LMArena Text: math | Not measured | 1494 | Only Kimi K3 measured |
| LMArena Text: mathematical work | Not measured | 1505 | Only Kimi K3 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index11 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 88.0% | Only Kimi K3 measured |
| LegalBench (Vals AI) | Not measured | 86.2% | Only Kimi K3 measured |
| CorpFin (Vals AI) | Not measured | 71.6% | Only Kimi K3 measured |
| TaxEval (Vals AI) | Not measured | 75.7% | Only Kimi K3 measured |
| MortgageTax (Vals AI) | Not measured | 66.3% | Only Kimi K3 measured |
| MedCode (Vals AI) | Not measured | 49.4% | Only Kimi K3 measured |
| SimpleQA Verified | Not measured | 50.6% | Only Kimi K3 measured |
| LMArena Text: science | Not measured | 1508 | Only Kimi K3 measured |
| LMArena Text: medicine and healthcare | Not measured | 1489 | Only Kimi K3 measured |
| LMArena Text: legal and government | Not measured | 1503 | Only Kimi K3 measured |
| LMArena Text: business and finance | Not measured | 1470 | Only Kimi K3 measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index8 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1334 | Only Kimi K3 measured |
| Design Arena: UI components | Not measured | 1341 | Only Kimi K3 measured |
| Design Arena: game dev | Not measured | 1358 | Only Kimi K3 measured |
| Design Arena: data viz | Not measured | 1351 | Only Kimi K3 measured |
| Design Arena: 3D | Not measured | 1387 | Only Kimi K3 measured |
| Design Arena: all code categories | Not measured | 1360 | Only Kimi K3 measured |
| Design Arena: SVG | Not measured | 1302 | Only Kimi K3 measured |
| LiveBench design | Not measured | 77.2% | Only Kimi K3 measured |
| Neither model has a result on the other 1 board here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index28 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | Leader |
|---|---|---|---|
| Vals Index | Not measured | 53.0% | Only Kimi K3 measured |
| LiveBench data analysis | Not measured | 78.7% | Only Kimi K3 measured |
| Public Benefits Bench (Vals AI) | Not measured | 68.2% | Only Kimi K3 measured |
| SAGE (Vals AI) | Not measured | 52.8% | Only Kimi K3 measured |
| MedScribe (Vals AI) | Not measured | 88.0% | Only Kimi K3 measured |
| Vals Multimodal Index | Not measured | 73.4% | Only Kimi K3 measured |
| Long context reasoning (AA-LCR) | Not measured | 88.7% | Only Kimi K3 measured |
| DTBench | Not measured | 91.2% | Only Kimi K3 measured |
| LMCA | Not measured | 52.7% | Only Kimi K3 measured |
| Surface Evolver Bench | Not measured | 95.0% | Only Kimi K3 measured |
| Furniture assembly | Not measured | 34.2% | Only Kimi K3 measured |
| LMArena Text | Not measured | 1476 | Only Kimi K3 measured |
| LMArena Text: instruction following | Not measured | 1485 | Only Kimi K3 measured |
| LMArena Text: creative writing | Not measured | 1457 | Only Kimi K3 measured |
| LMArena Text: English | Not measured | 1481 | Only Kimi K3 measured |
| LMArena Text: non-English | Not measured | 1466 | Only Kimi K3 measured |
| LMArena Text: Chinese | Not measured | 1529 | Only Kimi K3 measured |
| LMArena Text: German | Not measured | 1490 | Only Kimi K3 measured |
| LMArena Text: Japanese | Not measured | 1494 | Only Kimi K3 measured |
| LMArena Text: Korean | Not measured | 1460 | Only Kimi K3 measured |
| LMArena Text: French | Not measured | 1492 | Only Kimi K3 measured |
| LMArena Text: Spanish | Not measured | 1472 | Only Kimi K3 measured |
| LMArena Text: Russian | Not measured | 1485 | Only Kimi K3 measured |
| LMArena Text: Polish | Not measured | 1458 | Only Kimi K3 measured |
| LMArena Text: multi-turn | Not measured | 1489 | Only Kimi K3 measured |
| LMArena Text: longer queries | Not measured | 1494 | Only Kimi K3 measured |
| LMArena Text: writing and language | Not measured | 1471 | Only Kimi K3 measured |
| LMArena Text: entertainment and media | Not measured | 1453 | Only Kimi K3 measured |
| Neither model has a result on the other 18 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Devstral 2 2512 | Kimi K3 | 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
Devstral 2 2512 costs less
Repository review
50K input + 3K output tokens
Devstral 2 2512 costs less
Agent loop
200K input + 10K output tokens
Devstral 2 2512 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 | Mistral Devstral 2 2512 | MoonshotAI Kimi K3 |
|---|---|---|
| Maker | Mistral AI | Moonshot AI |
| Weights | Open weights | Open weights |
| Released | December 9, 2025 | July 16, 2026 |
| Context window | 262K | 1.05M |
| Input $/Mtok | $0.40 | $3.00 |
| Output $/Mtok | $2.00 | $15.00 |
| Coding index | Not benchmarked | 62.3 |
| Independent measurements behind it | 1 | 6 |
| Cost per solved task | $1.19 | $0.81 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Devstral 2 2512 better than Kimi K3 for coding?
Only Kimi K3 has a coding score, 62.3 on the Vibe Coding Index; Devstral 2 2512 has none, so the two cannot be ranked on coding. Pick Devstral 2 2512 for output tokens 7.5x cheaper ($2.00 vs $15.00 per million).
Which is cheaper, Devstral 2 2512 or Kimi K3?
Devstral 2 2512 costs $0.40 per million input tokens and $2.00 per million output tokens; Kimi K3 costs $3.00 and $15.00. Per solved SWE-bench Verified task, Devstral 2 2512 costs $1.19 and Kimi K3 $0.81.
Which has the bigger context window, Devstral 2 2512 or Kimi K3?
Kimi K3, with 1.05M tokens against 262K.
How many benchmarks is the Devstral 2 2512 vs Kimi K3 comparison based on?
5 benchmarks have a result for both models as of October 22, 2026. Devstral 2 2512's coding index rests on 1 independent benchmark and Kimi K3's on 6. Where only one model has a result, the page says so instead of naming a winner.