Devstral 2 2512 vs Grok 4.6
Grok 4.6 leads Devstral 2 2512 on the Artificial Analysis Intelligence Index, 44.3 to 8.6, and Devstral 2 2512 costs 3.0x less per output token ($2.00 vs $6.00 per million). Updated October 22, 2026; 5 benchmarks have a result for both.
Devstral 2 2512 or Grok 4.6: which to pick
Only Grok 4.6 has a coding score, 65.7 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 3.0x cheaper ($2.00 vs $6.00 per million).
- Shared results
- 5
- Devstral 2 2512 only
- 1
- Grok 4.6 only
- 118
- 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 Grok 4.6 has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
Grok 4.6
Grok 4.6 leads, 44.3 against 8.6.
Same board, both measured
Agentic work
AA Agentic Index
Grok 4.6
Grok 4.6 leads, 53.0 against 2.3.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Grok 4.6 has a GPQA Diamond result, 94.7%.
Directional only
Knowledge
MMLU Pro
No call
Only Grok 4.6 has a MMLU Pro result, 89.4%.
Directional only
Web design
Design Arena: websites
No call
Only Grok 4.6 has a Design Arena: websites result, 1289.
Directional only
Long documents
Listed context window
Grok 4.6
Grok 4.6 takes 500K tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Devstral 2 2512
Devstral 2 2512 $0.0014, Grok 4.6 $0.0050 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Grok 4.6
Devstral 2 2512 $1.19, Grok 4.6 $0.82.
Published attempt cost and pass rate
At a glance
- Coding index
- Devstral 2 2512 not benchmarked, Grok 4.6 65.7 from 6 benchmarks.
- Widest category
- Agentic work, 50.7 points apart on average over 1 shared benchmark.
- Widest single gap
- AA Agentic Index: Devstral 2 2512 2.3 against Grok 4.6 53.0, 50.7 points apart on a 0 to 100 scale.
- Output price
- Devstral 2 2512 $2.00 against Grok 4.6 $6.00 per million output tokens, 3.0x apart.
- Context window
- Devstral 2 2512 takes 262K tokens, Grok 4.6 500K.
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 | Grok 4.6 | 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 | Grok 4.6 | Leader |
|---|---|---|---|
| SWE-bench Verified | 53.8% | Not measured | Only Devstral 2 2512 measured |
| LiveBench coding | Not measured | 89.1% | Only Grok 4.6 measured |
| LMArena WebDev | Not measured | 1617 | Only Grok 4.6 measured |
| AA Coding Index | 31.3 | 76.8 | Grok 4.6 |
| LiveCodeBench (Vals AI) | 51.8% | 88.2% | Grok 4.6 |
| SWE-bench Verified (Vals AI) | 62.8% | 95.6% | Grok 4.6 |
| Terminal-Bench 4 (Vals AI) | Not measured | 17.2% | Only Grok 4.6 measured |
| Vibe Code Bench (Vals AI) | Not measured | 76.2% | Only Grok 4.6 measured |
| Code migration (Vals AI) | Not measured | 44.6% | Only Grok 4.6 measured |
| ProgramBench (Vals AI) | Not measured | 1.0% | Only Grok 4.6 measured |
| IOI (Vals AI) | Not measured | 47.6% | Only Grok 4.6 measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 78.3% | Only Grok 4.6 measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 14.8% | Only Grok 4.6 measured |
| SkillsBench (Vals AI) | Not measured | 55.8% | Only Grok 4.6 measured |
| CyberBench (Vals AI) | Not measured | 67.7% | Only Grok 4.6 measured |
| SciCode | Not measured | 56.5% | Only Grok 4.6 measured |
| FrontierCode | Not measured | 48.0% | Only Grok 4.6 measured |
| FrontierSWE | Not measured | 25.3% | Only Grok 4.6 measured |
| WeirdML | Not measured | 67.3% | Only Grok 4.6 measured |
| ALE-Bench | Not measured | 1508.1 | Only Grok 4.6 measured |
| LMArena Text: coding | Not measured | 1466 | Only Grok 4.6 measured |
| LMArena Text: software and IT | Not measured | 1457 | Only Grok 4.6 measured |
| LMArena WebDev: React | Not measured | 1631 | Only Grok 4.6 measured |
| LMArena WebDev: HTML | Not measured | 1597 | Only Grok 4.6 measured |
| LMArena WebDev: from an image | Not measured | 1603 | Only Grok 4.6 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 | Grok 4.6 | Leader |
|---|---|---|---|
| AA Intelligence Index | 8.6 | 44.3 | Grok 4.6 |
| Epoch Capabilities Index | Not measured | 156.6 | Only Grok 4.6 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Grok 4.6 | Leader |
|---|---|---|---|
| AA Agentic Index | 2.3 | 53.0 | Grok 4.6 |
| Design Arena: agents building web apps | Not measured | 1248 | Only Grok 4.6 measured |
| Design Arena: agents building full-stack apps | Not measured | 1264 | Only Grok 4.6 measured |
| Design Arena: agents building mobile apps | Not measured | 1234 | Only Grok 4.6 measured |
| Design Arena: agents building Android apps | Not measured | 1291 | Only Grok 4.6 measured |
| Design Arena: agents building Godot games | Not measured | 1276 | Only Grok 4.6 measured |
| Design Arena: agents building games | Not measured | 1209 | Only Grok 4.6 measured |
| Design Arena: HTML slides | Not measured | 1250 | Only Grok 4.6 measured |
| Design Arena: PowerPoint slides | Not measured | 1237 | Only Grok 4.6 measured |
| LiveBench research | Not measured | 84.5% | Only Grok 4.6 measured |
| LiveBench browser | Not measured | 64.0% | Only Grok 4.6 measured |
| LiveBench long agentic | Not measured | 40.7% | Only Grok 4.6 measured |
| Finance Agent (Vals AI) | Not measured | 53.7% | Only Grok 4.6 measured |
| Tax Agent Bench (Vals AI) | Not measured | 33.1% | Only Grok 4.6 measured |
| Legal Research Bench (Vals AI) | Not measured | 48.1% | Only Grok 4.6 measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 62.7% | Only Grok 4.6 measured |
| BioMysteryBench (Vals AI) | Not measured | 72.2% | Only Grok 4.6 measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 11.4% | Only Grok 4.6 measured |
| Vals Science Index | Not measured | 38.1% | Only Grok 4.6 measured |
| Time Horizon Index (Vals AI) | Not measured | 5.8% | Only Grok 4.6 measured |
| APEX-Agents | Not measured | 65.3% | Only Grok 4.6 measured |
| Vending-Bench 2 | Not measured | $9,047 | Only Grok 4.6 measured |
| Neither model has a result on the other 19 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index17 boards, none shared
| Benchmark | Devstral 2 2512 | Grok 4.6 | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 94.7% | Only Grok 4.6 measured |
| LiveBench reasoning | Not measured | 91.2% | Only Grok 4.6 measured |
| MysteryMechanism (Vals AI) | Not measured | 30.6% | Only Grok 4.6 measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 94.9% | Only Grok 4.6 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 44.1% | Only Grok 4.6 measured |
| GPQA Diamond (Epoch AI run) | Not measured | 94.0% | Only Grok 4.6 measured |
| ARC-AGI | Not measured | 87.5% | Only Grok 4.6 measured |
| ARC-AGI-2 | Not measured | 67.1% | Only Grok 4.6 measured |
| SimpleBench | Not measured | 75.9% | Only Grok 4.6 measured |
| Chess puzzles | Not measured | 33.0% | Only Grok 4.6 measured |
| Mystery game puzzles | Not measured | 34.0% | Only Grok 4.6 measured |
| EBR-bench | Not measured | 30.5% | Only Grok 4.6 measured |
| LMArena Text: hard prompts | Not measured | 1448 | Only Grok 4.6 measured |
| LMArena Text: expert | Not measured | 1469 | Only Grok 4.6 measured |
| LMArena Text: hard prompts in English | Not measured | 1445 | Only Grok 4.6 measured |
| LMArena Vision: diagrams | Not measured | 1285 | Only Grok 4.6 measured |
| LMArena Vision: homework | Not measured | 1283 | Only Grok 4.6 measured |
| Neither model has a result on the other 5 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Devstral 2 2512 | Grok 4.6 | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 99.2% | Only Grok 4.6 measured |
| FrontierMath, tiers 1 to 3 | Not measured | 66.0% | Only Grok 4.6 measured |
| FrontierMath, tier 4 | Not measured | 31.7% | Only Grok 4.6 measured |
| ProofBench | Not measured | 51.0% | Only Grok 4.6 measured |
| LMArena Text: math | Not measured | 1417 | Only Grok 4.6 measured |
| LMArena Text: mathematical work | Not measured | 1464 | Only Grok 4.6 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 | Grok 4.6 | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 89.4% | Only Grok 4.6 measured |
| LegalBench (Vals AI) | Not measured | 86.3% | Only Grok 4.6 measured |
| CorpFin (Vals AI) | Not measured | 66.2% | Only Grok 4.6 measured |
| TaxEval (Vals AI) | Not measured | 71.1% | Only Grok 4.6 measured |
| MortgageTax (Vals AI) | Not measured | 64.2% | Only Grok 4.6 measured |
| MedCode (Vals AI) | Not measured | 44.7% | Only Grok 4.6 measured |
| SimpleQA Verified | Not measured | 49.3% | Only Grok 4.6 measured |
| LMArena Text: science | Not measured | 1437 | Only Grok 4.6 measured |
| LMArena Text: medicine and healthcare | Not measured | 1420 | Only Grok 4.6 measured |
| LMArena Text: legal and government | Not measured | 1429 | Only Grok 4.6 measured |
| LMArena Text: business and finance | Not measured | 1416 | Only Grok 4.6 measured |
| Neither model has a result on the other 7 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index9 boards, none shared
| Benchmark | Devstral 2 2512 | Grok 4.6 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1289 | Only Grok 4.6 measured |
| Design Arena: UI components | Not measured | 1294 | Only Grok 4.6 measured |
| Design Arena: game dev | Not measured | 1296 | Only Grok 4.6 measured |
| Design Arena: data viz | Not measured | 1282 | Only Grok 4.6 measured |
| Design Arena: 3D | Not measured | 1277 | Only Grok 4.6 measured |
| Design Arena: all code categories | Not measured | 1292 | Only Grok 4.6 measured |
| Design Arena: SVG | Not measured | 1240 | Only Grok 4.6 measured |
| Design Arena: ASCII art | Not measured | 1279 | Only Grok 4.6 measured |
| LiveBench design | Not measured | 79.8% | Only Grok 4.6 measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index32 boards, none shared
| Benchmark | Devstral 2 2512 | Grok 4.6 | Leader |
|---|---|---|---|
| LiveBench data analysis | Not measured | 75.3% | Only Grok 4.6 measured |
| Public Benefits Bench (Vals AI) | Not measured | 66.8% | Only Grok 4.6 measured |
| SAGE (Vals AI) | Not measured | 28.9% | Only Grok 4.6 measured |
| MedScribe (Vals AI) | Not measured | 86.5% | Only Grok 4.6 measured |
| Long context reasoning (AA-LCR) | Not measured | 81.0% | Only Grok 4.6 measured |
| DTBench | Not measured | 97.3% | Only Grok 4.6 measured |
| LMCA | Not measured | 48.5% | Only Grok 4.6 measured |
| Furniture assembly | Not measured | 40.0% | Only Grok 4.6 measured |
| LMArena Text | Not measured | 1429 | Only Grok 4.6 measured |
| LMArena Text: instruction following | Not measured | 1431 | Only Grok 4.6 measured |
| LMArena Text: creative writing | Not measured | 1426 | Only Grok 4.6 measured |
| LMArena Vision | Not measured | 1263 | Only Grok 4.6 measured |
| LMArena Text: English | Not measured | 1433 | Only Grok 4.6 measured |
| LMArena Text: non-English | Not measured | 1420 | Only Grok 4.6 measured |
| LMArena Text: Chinese | Not measured | 1476 | Only Grok 4.6 measured |
| LMArena Text: German | Not measured | 1426 | Only Grok 4.6 measured |
| LMArena Text: Japanese | Not measured | 1378 | Only Grok 4.6 measured |
| LMArena Text: Korean | Not measured | 1391 | Only Grok 4.6 measured |
| LMArena Text: French | Not measured | 1455 | Only Grok 4.6 measured |
| LMArena Text: Spanish | Not measured | 1414 | Only Grok 4.6 measured |
| LMArena Text: Russian | Not measured | 1421 | Only Grok 4.6 measured |
| LMArena Text: Polish | Not measured | 1430 | Only Grok 4.6 measured |
| LMArena Text: multi-turn | Not measured | 1425 | Only Grok 4.6 measured |
| LMArena Text: longer queries | Not measured | 1456 | Only Grok 4.6 measured |
| LMArena Text: writing and language | Not measured | 1429 | Only Grok 4.6 measured |
| LMArena Text: entertainment and media | Not measured | 1421 | Only Grok 4.6 measured |
| LMArena Vision: English | Not measured | 1270 | Only Grok 4.6 measured |
| LMArena Vision: Chinese | Not measured | 1289 | Only Grok 4.6 measured |
| LMArena Vision: reading text | Not measured | 1272 | Only Grok 4.6 measured |
| LMArena Vision: humor | Not measured | 1270 | Only Grok 4.6 measured |
| LMArena Vision: creative writing | Not measured | 1262 | Only Grok 4.6 measured |
| LMArena Document | Not measured | 1452 | Only Grok 4.6 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 | Devstral 2 2512 | Grok 4.6 | 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 | SpaceXAI Grok 4.6 |
|---|---|---|
| Maker | Mistral AI | xAI |
| Weights | Open weights | No public weights listed |
| Released | December 9, 2025 | August 12, 2026 |
| Context window | 262K | 500K |
| Input $/Mtok | $0.40 | $2.00 |
| Output $/Mtok | $2.00 | $6.00 |
| Coding index | Not benchmarked | 65.7 |
| Independent measurements behind it | 1 | 6 |
| Cost per solved task | $1.19 | $0.82 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Devstral 2 2512 better than Grok 4.6 for coding?
Only Grok 4.6 has a coding score, 65.7 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 3.0x cheaper ($2.00 vs $6.00 per million).
Which is cheaper, Devstral 2 2512 or Grok 4.6?
Devstral 2 2512 costs $0.40 per million input tokens and $2.00 per million output tokens; Grok 4.6 costs $2.00 and $6.00. Per solved SWE-bench Verified task, Devstral 2 2512 costs $1.19 and Grok 4.6 $0.82.
Which has the bigger context window, Devstral 2 2512 or Grok 4.6?
Grok 4.6, with 500K tokens against 262K.
How many benchmarks is the Devstral 2 2512 vs Grok 4.6 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 Grok 4.6's on 6. Where only one model has a result, the page says so instead of naming a winner.