Claude Sonnet 4.6 vs Devstral 2 2512
Updated October 21, 2026. 5 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
vs boards won of 5 shared
Decision reading
Pick Claude Sonnet 4.6 for coding: it scores 57.8 on the Vibe Coding Index against 53.8 for Devstral 2 2512.
Pick Devstral 2 2512 when output tokens cost 7.5x less, or when a solved task costs $1.19 against $1.68.
Model AClaude Sonnet 4.6Anthropic57.8/1002 independent benchmarks$1.68 per solved taskModel BDevstral 2 2512Mistral AI53.8/1001 independent benchmark$1.19 per solved task
- Shared results
- 5
- Claude Sonnet 4.6 only
- 112
- Devstral 2 2512 only
- 1
- Like-for-like categories
- 3 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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, four public coding benchmarks
Claude Sonnet 4.6
Claude Sonnet 4.6 scores 57.8 against 53.8.
2 and 1 benchmarks behind the scores
Intelligence
AA Intelligence Index
Claude Sonnet 4.6
Claude Sonnet 4.6 leads, 30.1 against 8.6.
Same board, both measured
Agentic work
AA Agentic Index
Claude Sonnet 4.6
Claude Sonnet 4.6 leads, 31.8 against 2.3.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Claude Sonnet 4.6 has a GPQA Diamond result, 85.6%.
Directional only
Math
AIME
No call
Only Claude Sonnet 4.6 has a AIME result, 92.3%.
Directional only
Knowledge
MMLU Pro
No call
Only Claude Sonnet 4.6 has a MMLU Pro result, 87.3%.
Directional only
Web design
Design Arena: websites
No call
Only Claude Sonnet 4.6 has a Design Arena: websites result, 1285.
Directional only
Long documents
Listed context window
Claude Sonnet 4.6
Claude Sonnet 4.6 takes 1M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Devstral 2 2512
Claude Sonnet 4.6 $0.011, Devstral 2 2512 $0.0014 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Devstral 2 2512
Claude Sonnet 4.6 $1.68, Devstral 2 2512 $1.19.
Published attempt cost and pass rate
At a glance
- Coding index
- Claude Sonnet 4.6 57.8 from 2 benchmarks, Devstral 2 2512 53.8 from 1 benchmark.
- Widest category
- Agentic work, 29.5 points apart on average over 1 shared benchmark.
- Widest single gap
- AA Coding Index: Claude Sonnet 4.6 63.0 against Devstral 2 2512 31.3, 31.7 points apart on a 0 to 100 scale.
- Output price
- Claude Sonnet 4.6 $15.00 against Devstral 2 2512 $2.00 per million output tokens, 7.5x apart.
- Context window
- Claude Sonnet 4.6 takes 1M tokens, Devstral 2 2512 262K.
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 | Claude Sonnet 4.6 | Bars | Devstral 2 2512 | Ties |
|---|---|---|---|---|
| Coding | 3 | 0 | ||
| Intelligence | 1 | 0 | ||
| Agentic work | 1 | 0 | ||
| All shared boards | 5 | 0 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index30
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| SWE-bench Verified | Not measured | 53.8% | Only Devstral 2 2512 measured |
| LiveBench coding | 66.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev | 1523 | Not measured | Only Claude Sonnet 4.6 measured |
| AA Coding Index | 63.0 | 31.3 | Claude Sonnet 4.6 |
| LiveCodeBench (Vals AI) | 82.1% | 51.8% | Claude Sonnet 4.6 |
| SWE-bench Verified (Vals AI) | 77.4% | 62.8% | Claude Sonnet 4.6 |
| Vibe Code Bench (Vals AI) | 55.8% | Not measured | Only Claude Sonnet 4.6 measured |
| Code migration (Vals AI) | 39.9% | Not measured | Only Claude Sonnet 4.6 measured |
| ProgramBench (Vals AI) | 0.5% | Not measured | Only Claude Sonnet 4.6 measured |
| Terminal-Bench 2.1 (Vals AI) | 57.3% | Not measured | Only Claude Sonnet 4.6 measured |
| SkillsBench (Vals AI) | 49.0% | Not measured | Only Claude Sonnet 4.6 measured |
| SciCode | 50.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 53.0% | Not measured | Only Claude Sonnet 4.6 measured |
| Terminal-Bench | 53.4% | Not measured | Only Claude Sonnet 4.6 measured |
| FrontierCode | 24.3% | Not measured | Only Claude Sonnet 4.6 measured |
| WeirdML | 66.1% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: coding | 1503 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: software and IT | 1491 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev: React | 1512 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev: HTML | 1527 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena WebDev: from an image | 1529 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 17 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| AA Intelligence Index | 30.1 | 8.6 | Claude Sonnet 4.6 |
| Epoch Capabilities Index | 152.8 | Not measured | Only Claude Sonnet 4.6 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| AA Agentic Index | 31.8 | 2.3 | Claude Sonnet 4.6 |
| Design Arena: agents building web apps | 1190 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building full-stack apps | 1213 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building mobile apps | 1212 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building Android apps | 1202 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building Godot games | 1226 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: agents building games | 1177 | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench research | 63.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench browser | 68.0% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench long agentic | 23.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Finance Agent (Vals AI) | 51.0% | Not measured | Only Claude Sonnet 4.6 measured |
| Tax Agent Bench (Vals AI) | 25.8% | Not measured | Only Claude Sonnet 4.6 measured |
| Legal Research Bench (Vals AI) | 38.5% | Not measured | Only Claude Sonnet 4.6 measured |
| Excel Modeling Benchmark (Vals AI) | 60.2% | Not measured | Only Claude Sonnet 4.6 measured |
| τ²-Bench Telecom | 79.5% | Not measured | Only Claude Sonnet 4.6 measured |
| APEX-Agents | 43.0% | Not measured | Only Claude Sonnet 4.6 measured |
| OSWorld 2.0 | 9.3% | Not measured | Only Claude Sonnet 4.6 measured |
| DeepResearch Bench | 54.9% | Not measured | Only Claude Sonnet 4.6 measured |
| OSWorld | 72.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Vending-Bench 2 | $7,204 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 19 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index15 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| GPQA Diamond | 85.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench reasoning | 86.4% | Not measured | Only Claude Sonnet 4.6 measured |
| MMMU Pro (Vals AI) | 83.6% | Not measured | Only Claude Sonnet 4.6 measured |
| GPQA Diamond (Artificial Analysis run) | 87.5% | Not measured | Only Claude Sonnet 4.6 measured |
| Humanity's Last Exam (Artificial Analysis run) | 33.6% | Not measured | Only Claude Sonnet 4.6 measured |
| GPQA Diamond (Epoch AI run) | 87.4% | Not measured | Only Claude Sonnet 4.6 measured |
| ARC-AGI | 86.5% | Not measured | Only Claude Sonnet 4.6 measured |
| ARC-AGI-2 | 60.4% | Not measured | Only Claude Sonnet 4.6 measured |
| Chess puzzles | 23.0% | Not measured | Only Claude Sonnet 4.6 measured |
| Mystery game puzzles | 16.0% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: hard prompts | 1485 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: expert | 1500 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: hard prompts in English | 1493 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: diagrams | 1306 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: homework | 1311 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 6 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| AIME | 92.3% | Not measured | Only Claude Sonnet 4.6 measured |
| OTIS Mock AIME | 85.8% | Not measured | Only Claude Sonnet 4.6 measured |
| ProofBench | 45.0% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: math | 1462 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: mathematical work | 1483 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 8 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index12 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| MMLU Pro | 87.3% | Not measured | Only Claude Sonnet 4.6 measured |
| LegalBench (Vals AI) | 82.1% | Not measured | Only Claude Sonnet 4.6 measured |
| CaseLaw (Vals AI) | 64.0% | Not measured | Only Claude Sonnet 4.6 measured |
| CorpFin (Vals AI) | 65.3% | Not measured | Only Claude Sonnet 4.6 measured |
| TaxEval (Vals AI) | 77.1% | Not measured | Only Claude Sonnet 4.6 measured |
| MortgageTax (Vals AI) | 67.7% | Not measured | Only Claude Sonnet 4.6 measured |
| MedQA (Vals AI) | 92.1% | Not measured | Only Claude Sonnet 4.6 measured |
| SimpleQA Verified | 35.5% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: science | 1474 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: medicine and healthcare | 1451 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: legal and government | 1470 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: business and finance | 1464 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| Design Arena: websites | 1285 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: UI components | 1271 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: game dev | 1259 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: data viz | 1278 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: 3D | 1234 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: all code categories | 1276 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: SVG | 1196 | Not measured | Only Claude Sonnet 4.6 measured |
| Design Arena: ASCII art | 1235 | Not measured | Only Claude Sonnet 4.6 measured |
| LiveBench design | 71.8% | Not measured | Only Claude Sonnet 4.6 measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index34 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | Leader |
|---|---|---|---|
| LiveBench data analysis | 76.1% | Not measured | Only Claude Sonnet 4.6 measured |
| Public Benefits Bench (Vals AI) | 62.5% | Not measured | Only Claude Sonnet 4.6 measured |
| SAGE (Vals AI) | 46.6% | Not measured | Only Claude Sonnet 4.6 measured |
| Vals Multimodal Index | 60.6% | Not measured | Only Claude Sonnet 4.6 measured |
| IFBench | 56.6% | Not measured | Only Claude Sonnet 4.6 measured |
| Long context reasoning (AA-LCR) | 80.0% | Not measured | Only Claude Sonnet 4.6 measured |
| DTBench | 89.9% | Not measured | Only Claude Sonnet 4.6 measured |
| LMCA | 46.5% | Not measured | Only Claude Sonnet 4.6 measured |
| ExploitBench | 23.6% | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text | 1458 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: instruction following | 1476 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: creative writing | 1435 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision | 1283 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: English | 1472 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: non-English | 1440 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Chinese | 1491 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: German | 1428 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Japanese | 1420 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Korean | 1410 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: French | 1465 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Spanish | 1464 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Russian | 1440 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: Polish | 1443 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: multi-turn | 1464 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: longer queries | 1479 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: writing and language | 1444 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Text: entertainment and media | 1431 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: English | 1298 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: Chinese | 1307 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: reading text | 1296 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: humor | 1248 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: recognizing things | 1195 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Vision: creative writing | 1272 | Not measured | Only Claude Sonnet 4.6 measured |
| LMArena Document | 1482 | Not measured | Only Claude Sonnet 4.6 measured |
| Neither model has a result on the other 12 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Claude Sonnet 4.6 | Devstral 2 2512 | 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
Claude Sonnet 4.6$0.011Fits in one request
Devstral 2 2512$0.0014Fits in one request
Devstral 2 2512 costs less
Repository review
50K input + 3K output tokens
Claude Sonnet 4.6$0.195Fits in one request
Devstral 2 2512$0.026Fits in one request
Devstral 2 2512 costs less
Agent loop
200K input + 10K output tokens
Claude Sonnet 4.6$0.750Fits in one request
Devstral 2 2512$0.100Fits in one request
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 | Anthropic Claude Sonnet 4.6 | Mistral Devstral 2 2512 |
|---|---|---|
| Maker | Anthropic | Mistral AI |
| Weights | No public weights listed | Open weights |
| Released | February 17, 2026 | December 9, 2025 |
| Context window | 1M | 262K |
| Input $/Mtok | $3.00 | $0.40 |
| Output $/Mtok | $15.00 | $2.00 |
| Coding index | 57.8 | 53.8 |
| Independent measurements behind it | 2 | 1 |
| Cost per solved task | $1.68 | $1.19 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Claude Sonnet 4.6 better than Devstral 2 2512 for coding?
Pick Claude Sonnet 4.6 for coding: it scores 57.8 on the Vibe Coding Index against 53.8 for Devstral 2 2512. Pick Devstral 2 2512 when output tokens cost 7.5x less, or when a solved task costs $1.19 against $1.68.
Which is cheaper, Claude Sonnet 4.6 or Devstral 2 2512?
Claude Sonnet 4.6 costs $3.00 per million input tokens and $15.00 per million output tokens; Devstral 2 2512 costs $0.40 and $2.00. Per solved SWE-bench Verified task, Claude Sonnet 4.6 costs $1.68 and Devstral 2 2512 $1.19.
Which has the bigger context window, Claude Sonnet 4.6 or Devstral 2 2512?
Claude Sonnet 4.6, with 1M tokens against 262K.
How many benchmarks is the Claude Sonnet 4.6 vs Devstral 2 2512 comparison based on?
5 benchmarks have a result for both models as of October 21, 2026. Claude Sonnet 4.6's coding index rests on 2 independent benchmarks and Devstral 2 2512's on 1. Where only one model has a result, the page says so instead of naming a winner.