Devstral 2 2512 vs Gemini 3.8 Flash
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 Gemini 3.8 Flash for coding: it scores 70.5 on the Vibe Coding Index against 53.8 for Devstral 2 2512.
Pick Devstral 2 2512 when output tokens cost 2.1x less, or when a solved task costs $1.19 against $2.74.
Model ADevstral 2 2512Mistral AI53.8/1001 independent benchmark$1.19 per solved taskModel BGemini 3.8 FlashGoogle70.5/1002 independent benchmarks$2.74 per solved task
- Shared results
- 5
- Devstral 2 2512 only
- 1
- Gemini 3.8 Flash only
- 121
- 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
Gemini 3.8 Flash
Gemini 3.8 Flash scores 70.5 against 53.8.
1 and 2 benchmarks behind the scores
Intelligence
AA Intelligence Index
Gemini 3.8 Flash
Gemini 3.8 Flash leads, 40.9 against 8.6.
Same board, both measured
Agentic work
AA Agentic Index
Gemini 3.8 Flash
Gemini 3.8 Flash leads, 40.2 against 2.3.
Same board, both measured
Reasoning
GPQA Diamond
No call
Only Gemini 3.8 Flash has a GPQA Diamond result, 94.4%.
Directional only
Knowledge
MMLU Pro
No call
Only Gemini 3.8 Flash has a MMLU Pro result, 90.2%.
Directional only
Web design
Design Arena: websites
No call
Only Gemini 3.8 Flash has a Design Arena: websites result, 1301.
Directional only
General
Vals Index
No call
Only Gemini 3.8 Flash has a Vals Index result, 56.0%.
Directional only
Long documents
Listed context window
Gemini 3.8 Flash
Gemini 3.8 Flash takes 1.05M tokens against 262K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Devstral 2 2512
Devstral 2 2512 $0.0014, Gemini 3.8 Flash $0.0029 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
Devstral 2 2512
Devstral 2 2512 $1.19, Gemini 3.8 Flash $2.74.
Published attempt cost and pass rate
At a glance
- Coding index
- Devstral 2 2512 53.8 from 1 benchmark, Gemini 3.8 Flash 70.5 from 2 benchmarks.
- Widest category
- Agentic work, 37.9 points apart on average over 1 shared benchmark.
- Widest single gap
- AA Coding Index: Devstral 2 2512 31.3 against Gemini 3.8 Flash 76.3, 45.0 points apart on a 0 to 100 scale.
- Output price
- Devstral 2 2512 $2.00 against Gemini 3.8 Flash $4.125 per million output tokens, 2.1x apart.
- Context window
- Devstral 2 2512 takes 262K tokens, Gemini 3.8 Flash 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 | Gemini 3.8 Flash | Ties |
|---|---|---|---|---|
| Coding | 0 | 3 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 1 | ||
| All shared boards | 0 | 5 |
CodingLike-for-likeFour of these boards feed the Vibe Coding Index03
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| SWE-bench Verified | 53.8% | Not measured | Only Devstral 2 2512 measured |
| LiveBench coding | Not measured | 82.2% | Only Gemini 3.8 Flash measured |
| LMArena WebDev | Not measured | 1585 | Only Gemini 3.8 Flash measured |
| AA Coding Index | 31.3 | 76.3 | Gemini 3.8 Flash |
| LiveCodeBench (Vals AI) | 51.8% | 89.5% | Gemini 3.8 Flash |
| SWE-bench Verified (Vals AI) | 62.8% | 80.0% | Gemini 3.8 Flash |
| Terminal-Bench 4 (Vals AI) | Not measured | 19.7% | Only Gemini 3.8 Flash measured |
| Vibe Code Bench (Vals AI) | Not measured | 78.7% | Only Gemini 3.8 Flash measured |
| Code migration (Vals AI) | Not measured | 36.5% | Only Gemini 3.8 Flash measured |
| ProgramBench (Vals AI) | Not measured | 1.0% | Only Gemini 3.8 Flash measured |
| IOI (Vals AI) | Not measured | 56.9% | Only Gemini 3.8 Flash measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 81.3% | Only Gemini 3.8 Flash measured |
| SWE-bench Multimodal (Vals AI) | Not measured | 40.8% | Only Gemini 3.8 Flash measured |
| Vibe Code Bench 1-100 (Vals AI) | Not measured | 18.8% | Only Gemini 3.8 Flash measured |
| SkillsBench (Vals AI) | Not measured | 58.0% | Only Gemini 3.8 Flash measured |
| CyberBench (Vals AI) | Not measured | 43.8% | Only Gemini 3.8 Flash measured |
| SciCode | Not measured | 56.6% | Only Gemini 3.8 Flash measured |
| FrontierCode | Not measured | 41.2% | Only Gemini 3.8 Flash measured |
| FrontierSWE | Not measured | 19.6% | Only Gemini 3.8 Flash measured |
| WeirdML | Not measured | 84.8% | Only Gemini 3.8 Flash measured |
| ALE-Bench | Not measured | 1269.7 | Only Gemini 3.8 Flash measured |
| LMArena Text: coding | Not measured | 1508 | Only Gemini 3.8 Flash measured |
| LMArena Text: software and IT | Not measured | 1507 | Only Gemini 3.8 Flash measured |
| LMArena WebDev: React | Not measured | 1577 | Only Gemini 3.8 Flash measured |
| LMArena WebDev: HTML | Not measured | 1616 | Only Gemini 3.8 Flash measured |
| LMArena WebDev: from an image | Not measured | 1593 | Only Gemini 3.8 Flash measured |
| Neither model has a result on the other 12 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| AA Intelligence Index | 8.6 | 40.9 | Gemini 3.8 Flash |
| Epoch Capabilities Index | Not measured | 157.4 | Only Gemini 3.8 Flash measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| AA Agentic Index | 2.3 | 40.2 | Gemini 3.8 Flash |
| Design Arena: agents building web apps | Not measured | 1238 | Only Gemini 3.8 Flash measured |
| Design Arena: agents building full-stack apps | Not measured | 1244 | Only Gemini 3.8 Flash measured |
| Design Arena: agents building mobile apps | Not measured | 1236 | Only Gemini 3.8 Flash measured |
| Design Arena: agents building Android apps | Not measured | 1078 | Only Gemini 3.8 Flash measured |
| Design Arena: agents building Godot games | Not measured | 1273 | Only Gemini 3.8 Flash measured |
| Design Arena: PowerPoint slides | Not measured | 1178 | Only Gemini 3.8 Flash measured |
| Design Arena: spreadsheets | Not measured | 1052 | Only Gemini 3.8 Flash measured |
| LiveBench research | Not measured | 79.1% | Only Gemini 3.8 Flash measured |
| LiveBench browser | Not measured | 64.0% | Only Gemini 3.8 Flash measured |
| LiveBench long agentic | Not measured | 30.1% | Only Gemini 3.8 Flash measured |
| SREBench (Vals AI) | Not measured | 11.1% | Only Gemini 3.8 Flash measured |
| Finance Agent (Vals AI) | Not measured | 61.4% | Only Gemini 3.8 Flash measured |
| Tax Agent Bench (Vals AI) | Not measured | 32.1% | Only Gemini 3.8 Flash measured |
| Legal Research Bench (Vals AI) | Not measured | 38.9% | Only Gemini 3.8 Flash measured |
| Harvey's Legal Agent Benchmark (Vals AI) | Not measured | 12.1% | Only Gemini 3.8 Flash measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 72.2% | Only Gemini 3.8 Flash measured |
| BioMysteryBench (Vals AI) | Not measured | 62.2% | Only Gemini 3.8 Flash measured |
| Terminal-Bench-Science (Vals AI) | Not measured | 8.6% | Only Gemini 3.8 Flash measured |
| Vals Science Index | Not measured | 35.8% | Only Gemini 3.8 Flash measured |
| Time Horizon Index (Vals AI) | Not measured | 18.8% | Only Gemini 3.8 Flash measured |
| CUA-bench (Vals AI) | Not measured | 4.2% | Only Gemini 3.8 Flash measured |
| APEX-Agents | Not measured | 64.3% | Only Gemini 3.8 Flash measured |
| Remote Labor Index | Not measured | 5.8% | Only Gemini 3.8 Flash measured |
| Vending-Bench 2 | Not measured | $5,094 | Only Gemini 3.8 Flash measured |
| Neither model has a result on the other 14 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index16 boards, none shared
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 94.4% | Only Gemini 3.8 Flash measured |
| LiveBench reasoning | Not measured | 89.3% | Only Gemini 3.8 Flash measured |
| MysteryMechanism (Vals AI) | Not measured | 36.5% | Only Gemini 3.8 Flash measured |
| MMMU Pro (Vals AI) | Not measured | 89.1% | Only Gemini 3.8 Flash measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 95.3% | Only Gemini 3.8 Flash measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 47.8% | Only Gemini 3.8 Flash measured |
| GPQA Diamond (Epoch AI run) | Not measured | 95.4% | Only Gemini 3.8 Flash measured |
| ARC-AGI | Not measured | 98.5% | Only Gemini 3.8 Flash measured |
| ARC-AGI-2 | Not measured | 89.2% | Only Gemini 3.8 Flash measured |
| Chess puzzles | Not measured | 58.0% | Only Gemini 3.8 Flash measured |
| Mystery game puzzles | Not measured | 47.0% | Only Gemini 3.8 Flash measured |
| LMArena Text: hard prompts | Not measured | 1509 | Only Gemini 3.8 Flash measured |
| LMArena Text: expert | Not measured | 1526 | Only Gemini 3.8 Flash measured |
| LMArena Text: hard prompts in English | Not measured | 1504 | Only Gemini 3.8 Flash measured |
| LMArena Vision: diagrams | Not measured | 1345 | Only Gemini 3.8 Flash measured |
| LMArena Vision: homework | Not measured | 1351 | Only Gemini 3.8 Flash 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 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| OTIS Mock AIME | Not measured | 98.9% | Only Gemini 3.8 Flash measured |
| FrontierMath, tiers 1 to 3 | Not measured | 68.4% | Only Gemini 3.8 Flash measured |
| FrontierMath, tier 4 | Not measured | 22.0% | Only Gemini 3.8 Flash measured |
| ProofBench | Not measured | 48.0% | Only Gemini 3.8 Flash measured |
| LMArena Text: math | Not measured | 1532 | Only Gemini 3.8 Flash measured |
| LMArena Text: mathematical work | Not measured | 1519 | Only Gemini 3.8 Flash measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index10 boards, none shared
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 90.2% | Only Gemini 3.8 Flash measured |
| LegalBench (Vals AI) | Not measured | 87.0% | Only Gemini 3.8 Flash measured |
| TaxEval (Vals AI) | Not measured | 74.5% | Only Gemini 3.8 Flash measured |
| MortgageTax (Vals AI) | Not measured | 65.3% | Only Gemini 3.8 Flash measured |
| MedCode (Vals AI) | Not measured | 48.1% | Only Gemini 3.8 Flash measured |
| SimpleQA Verified | Not measured | 69.7% | Only Gemini 3.8 Flash measured |
| LMArena Text: science | Not measured | 1523 | Only Gemini 3.8 Flash measured |
| LMArena Text: medicine and healthcare | Not measured | 1508 | Only Gemini 3.8 Flash measured |
| LMArena Text: legal and government | Not measured | 1497 | Only Gemini 3.8 Flash measured |
| LMArena Text: business and finance | Not measured | 1479 | Only Gemini 3.8 Flash 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 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1301 | Only Gemini 3.8 Flash measured |
| Design Arena: UI components | Not measured | 1318 | Only Gemini 3.8 Flash measured |
| Design Arena: game dev | Not measured | 1305 | Only Gemini 3.8 Flash measured |
| Design Arena: data viz | Not measured | 1269 | Only Gemini 3.8 Flash measured |
| Design Arena: 3D | Not measured | 1295 | Only Gemini 3.8 Flash measured |
| Design Arena: all code categories | Not measured | 1304 | Only Gemini 3.8 Flash measured |
| Design Arena: SVG | Not measured | 1295 | Only Gemini 3.8 Flash measured |
| Design Arena: ASCII art | Not measured | 1264 | Only Gemini 3.8 Flash measured |
| LiveBench design | Not measured | 72.7% | Only Gemini 3.8 Flash measured |
GeneralDirectional onlyPublic boards, shown as published, kept out of the index33 boards, none shared
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | Leader |
|---|---|---|---|
| Vals Index | Not measured | 56.0% | Only Gemini 3.8 Flash measured |
| LiveBench data analysis | Not measured | 54.0% | Only Gemini 3.8 Flash measured |
| Public Benefits Bench (Vals AI) | Not measured | 65.3% | Only Gemini 3.8 Flash measured |
| SAGE (Vals AI) | Not measured | 35.1% | Only Gemini 3.8 Flash measured |
| MedScribe (Vals AI) | Not measured | 84.5% | Only Gemini 3.8 Flash measured |
| Long context reasoning (AA-LCR) | Not measured | 84.0% | Only Gemini 3.8 Flash measured |
| DTBench | Not measured | 95.7% | Only Gemini 3.8 Flash measured |
| LMCA | Not measured | 52.9% | Only Gemini 3.8 Flash measured |
| Surface Evolver Bench | Not measured | 76.9% | Only Gemini 3.8 Flash measured |
| Furniture assembly | Not measured | 31.7% | Only Gemini 3.8 Flash measured |
| LMArena Text | Not measured | 1499 | Only Gemini 3.8 Flash measured |
| LMArena Text: instruction following | Not measured | 1492 | Only Gemini 3.8 Flash measured |
| LMArena Text: creative writing | Not measured | 1488 | Only Gemini 3.8 Flash measured |
| LMArena Vision | Not measured | 1315 | Only Gemini 3.8 Flash measured |
| LMArena Text: English | Not measured | 1498 | Only Gemini 3.8 Flash measured |
| LMArena Text: non-English | Not measured | 1491 | Only Gemini 3.8 Flash measured |
| LMArena Text: Chinese | Not measured | 1553 | Only Gemini 3.8 Flash measured |
| LMArena Text: German | Not measured | 1501 | Only Gemini 3.8 Flash measured |
| LMArena Text: Japanese | Not measured | 1507 | Only Gemini 3.8 Flash measured |
| LMArena Text: Korean | Not measured | 1464 | Only Gemini 3.8 Flash measured |
| LMArena Text: French | Not measured | 1496 | Only Gemini 3.8 Flash measured |
| LMArena Text: Spanish | Not measured | 1487 | Only Gemini 3.8 Flash measured |
| LMArena Text: Russian | Not measured | 1514 | Only Gemini 3.8 Flash measured |
| LMArena Text: Polish | Not measured | 1497 | Only Gemini 3.8 Flash measured |
| LMArena Text: multi-turn | Not measured | 1503 | Only Gemini 3.8 Flash measured |
| LMArena Text: longer queries | Not measured | 1507 | Only Gemini 3.8 Flash measured |
| LMArena Text: writing and language | Not measured | 1494 | Only Gemini 3.8 Flash measured |
| LMArena Text: entertainment and media | Not measured | 1480 | Only Gemini 3.8 Flash measured |
| LMArena Vision: English | Not measured | 1318 | Only Gemini 3.8 Flash measured |
| LMArena Vision: Chinese | Not measured | 1416 | Only Gemini 3.8 Flash measured |
| LMArena Vision: reading text | Not measured | 1333 | Only Gemini 3.8 Flash measured |
| LMArena Vision: humor | Not measured | 1321 | Only Gemini 3.8 Flash measured |
| LMArena Vision: creative writing | Not measured | 1299 | Only Gemini 3.8 Flash measured |
| Neither model has a result on the other 13 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Devstral 2 2512 | Gemini 3.8 Flash | 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$0.0014Fits in one request
Gemini 3.8 Flash$0.0029Fits in one request
Devstral 2 2512 costs less
Repository review
50K input + 3K output tokens
Devstral 2 2512$0.026Fits in one request
Gemini 3.8 Flash$0.054Fits in one request
Devstral 2 2512 costs less
Agent loop
200K input + 10K output tokens
Devstral 2 2512$0.100Fits in one request
Gemini 3.8 Flash$0.206Fits 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 | Mistral Devstral 2 2512 | Google Gemini 3.8 Flash |
|---|---|---|
| Maker | Mistral AI | |
| Weights | Open weights | No public weights listed |
| Released | December 9, 2025 | September 2, 2026 |
| Context window | 262K | 1.05M |
| Input $/Mtok | $0.40 | $0.825 |
| Output $/Mtok | $2.00 | $4.125 |
| Coding index | 53.8 | 70.5 |
| Independent measurements behind it | 1 | 2 |
| Cost per solved task | $1.19 | $2.74 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Devstral 2 2512 better than Gemini 3.8 Flash for coding?
Pick Gemini 3.8 Flash for coding: it scores 70.5 on the Vibe Coding Index against 53.8 for Devstral 2 2512. Pick Devstral 2 2512 when output tokens cost 2.1x less, or when a solved task costs $1.19 against $2.74.
Which is cheaper, Devstral 2 2512 or Gemini 3.8 Flash?
Devstral 2 2512 costs $0.40 per million input tokens and $2.00 per million output tokens; Gemini 3.8 Flash costs $0.825 and $4.125. Per solved SWE-bench Verified task, Devstral 2 2512 costs $1.19 and Gemini 3.8 Flash $2.74.
Which has the bigger context window, Devstral 2 2512 or Gemini 3.8 Flash?
Gemini 3.8 Flash, with 1.05M tokens against 262K.
How many benchmarks is the Devstral 2 2512 vs Gemini 3.8 Flash comparison based on?
5 benchmarks have a result for both models as of October 21, 2026. Devstral 2 2512's coding index rests on 1 independent benchmark and Gemini 3.8 Flash's on 2. Where only one model has a result, the page says so instead of naming a winner.