Codestral 2508 vs GPT-5.4 Nano
Updated October 21, 2026. 1 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Only GPT-5.4 Nano has a coding score, 44.8 on the Vibe Coding Index; Codestral 2508 has none, so the two cannot be ranked on coding.
Codestral 2508 has the edge where output tokens cost 1.4x less.
- Shared results
- 1
- Codestral 2508 only
- 6
- GPT-5.4 Nano only
- 83
- Like-for-like categories
- 1 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, six public coding benchmarks
No call
Only GPT-5.4 Nano has a coding score, so no call is made.
One side not measured
Intelligence
AA Intelligence Index
No call
Only GPT-5.4 Nano has a AA Intelligence Index result, 20.7.
Directional only
Agentic work
AA Agentic Index
No call
Only GPT-5.4 Nano has a AA Agentic Index result, 16.0.
Directional only
Reasoning
GPQA Diamond
No call
Only GPT-5.4 Nano has a GPQA Diamond result, 77.5%.
Directional only
Math
AIME
No call
Only GPT-5.4 Nano has a AIME result, 88.8%.
Directional only
Knowledge
MMLU Pro
No call
Only GPT-5.4 Nano has a MMLU Pro result, 77.2%.
Directional only
Web design
Design Arena: websites
No call
Only Codestral 2508 has a Design Arena: websites result, 1013.
Directional only
Long documents
Listed context window
GPT-5.4 Nano
GPT-5.4 Nano takes 400K tokens against 256K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Codestral 2508
Codestral 2508 $0.0008, GPT-5.4 Nano $0.0008 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only GPT-5.4 Nano has the number, $0.14.
Directional only
At a glance
- Coding index
- Codestral 2508 not benchmarked, GPT-5.4 Nano 44.8 from 2 benchmarks.
- Widest category
- Coding, 28.9 points apart on average over 1 shared benchmark.
- Widest single gap
- ALE-Bench: Codestral 2508 137.8 against GPT-5.4 Nano 1004.5, 28.9 points apart on a 0 to 100 scale.
- Output price
- Codestral 2508 $0.90 against GPT-5.4 Nano $1.25 per million output tokens, 1.4x apart.
- Context window
- Codestral 2508 takes 256K tokens, GPT-5.4 Nano 400K.
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 | Codestral 2508 | Bars | GPT-5.4 Nano | Ties |
|---|---|---|---|---|
| Coding | 0 | 1 | ||
| All shared boards | 0 | 1 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index01
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| AA Coding Index | Not measured | 56.1 | Only GPT-5.4 Nano measured |
| LiveCodeBench (Vals AI) | Not measured | 84.0% | Only GPT-5.4 Nano measured |
| SWE-bench Verified (Vals AI) | Not measured | 69.8% | Only GPT-5.4 Nano measured |
| Vibe Code Bench (Vals AI) | Not measured | 26.1% | Only GPT-5.4 Nano measured |
| Code migration (Vals AI) | Not measured | 14.5% | Only GPT-5.4 Nano measured |
| Terminal-Bench 2.1 (Vals AI) | Not measured | 41.6% | Only GPT-5.4 Nano measured |
| SciCode | Not measured | 47.2% | Only GPT-5.4 Nano measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 42.4% | Only GPT-5.4 Nano measured |
| WeirdML | Not measured | 49.2% | Only GPT-5.4 Nano measured |
| ALE-Bench | 137.8 | 1004.5 | GPT-5.4 Nano |
| LMArena Text: coding | Not measured | 1404 | Only GPT-5.4 Nano measured |
| LMArena Text: software and IT | Not measured | 1402 | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 26 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 20.7 | Only GPT-5.4 Nano measured |
| Epoch Capabilities Index | Not measured | 146.0 | Only GPT-5.4 Nano measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| AA Agentic Index | Not measured | 16.0 | Only GPT-5.4 Nano measured |
| Finance Agent (Vals AI) | Not measured | 38.2% | Only GPT-5.4 Nano measured |
| Tax Agent Bench (Vals AI) | Not measured | 5.0% | Only GPT-5.4 Nano measured |
| Legal Research Bench (Vals AI) | Not measured | 6.3% | Only GPT-5.4 Nano measured |
| Excel Modeling Benchmark (Vals AI) | Not measured | 44.8% | Only GPT-5.4 Nano measured |
| τ²-Bench Telecom | Not measured | 76.0% | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 33 boards here | |||
ReasoningDirectional onlyPublic boards, shown as published, kept out of the index14 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| GPQA Diamond | Not measured | 77.5% | Only GPT-5.4 Nano measured |
| MMMU Pro (Vals AI) | Not measured | 73.6% | Only GPT-5.4 Nano measured |
| GPQA Diamond (Artificial Analysis run) | Not measured | 81.7% | Only GPT-5.4 Nano measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 28.3% | Only GPT-5.4 Nano measured |
| GPQA Diamond (Epoch AI run) | Not measured | 78.5% | Only GPT-5.4 Nano measured |
| ARC-AGI | Not measured | 51.5% | Only GPT-5.4 Nano measured |
| ARC-AGI-2 | Not measured | 5.7% | Only GPT-5.4 Nano measured |
| Chess puzzles | Not measured | 45.0% | Only GPT-5.4 Nano measured |
| Mystery game puzzles | Not measured | 9.0% | Only GPT-5.4 Nano measured |
| LMArena Text: hard prompts | Not measured | 1381 | Only GPT-5.4 Nano measured |
| LMArena Text: expert | Not measured | 1396 | Only GPT-5.4 Nano measured |
| LMArena Text: hard prompts in English | Not measured | 1387 | Only GPT-5.4 Nano measured |
| LMArena Vision: diagrams | Not measured | 1210 | Only GPT-5.4 Nano measured |
| LMArena Vision: homework | Not measured | 1225 | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 7 boards here | |||
MathDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| AIME | Not measured | 88.8% | Only GPT-5.4 Nano measured |
| OTIS Mock AIME | Not measured | 87.8% | Only GPT-5.4 Nano measured |
| FrontierMath, tiers 1 to 3 | Not measured | 44.9% | Only GPT-5.4 Nano measured |
| FrontierMath, tier 4 | Not measured | 12.2% | Only GPT-5.4 Nano measured |
| ProofBench | Not measured | 5.0% | Only GPT-5.4 Nano measured |
| LMArena Text: math | Not measured | 1405 | Only GPT-5.4 Nano measured |
| LMArena Text: mathematical work | Not measured | 1409 | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 6 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index12 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| MMLU Pro | Not measured | 77.2% | Only GPT-5.4 Nano measured |
| LegalBench (Vals AI) | Not measured | 77.9% | Only GPT-5.4 Nano measured |
| CaseLaw (Vals AI) | Not measured | 51.9% | Only GPT-5.4 Nano measured |
| CorpFin (Vals AI) | Not measured | 61.2% | Only GPT-5.4 Nano measured |
| TaxEval (Vals AI) | Not measured | 67.4% | Only GPT-5.4 Nano measured |
| MortgageTax (Vals AI) | Not measured | 59.1% | Only GPT-5.4 Nano measured |
| MedCode (Vals AI) | Not measured | 41.0% | Only GPT-5.4 Nano measured |
| SimpleQA Verified | Not measured | 11.7% | Only GPT-5.4 Nano measured |
| LMArena Text: science | Not measured | 1374 | Only GPT-5.4 Nano measured |
| LMArena Text: medicine and healthcare | Not measured | 1369 | Only GPT-5.4 Nano measured |
| LMArena Text: legal and government | Not measured | 1368 | Only GPT-5.4 Nano measured |
| LMArena Text: business and finance | Not measured | 1366 | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| Design Arena: websites | 1013 | Not measured | Only Codestral 2508 measured |
| Design Arena: UI components | 1017 | Not measured | Only Codestral 2508 measured |
| Design Arena: game dev | 968 | Not measured | Only Codestral 2508 measured |
| Design Arena: data viz | 1017 | Not measured | Only Codestral 2508 measured |
| Design Arena: 3D | 1020 | Not measured | Only Codestral 2508 measured |
| Design Arena: all code categories | 1007 | Not measured | Only Codestral 2508 measured |
| Neither model has a result on the other 3 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index31 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | Leader |
|---|---|---|---|
| SAGE (Vals AI) | Not measured | 38.1% | Only GPT-5.4 Nano measured |
| MedScribe (Vals AI) | Not measured | 77.1% | Only GPT-5.4 Nano measured |
| Vals Multimodal Index | Not measured | 47.6% | Only GPT-5.4 Nano measured |
| IFBench | Not measured | 75.9% | Only GPT-5.4 Nano measured |
| Long context reasoning (AA-LCR) | Not measured | 76.7% | Only GPT-5.4 Nano measured |
| DTBench | Not measured | 80.3% | Only GPT-5.4 Nano measured |
| LMCA | Not measured | 36.9% | Only GPT-5.4 Nano measured |
| LMArena Text | Not measured | 1372 | Only GPT-5.4 Nano measured |
| LMArena Text: instruction following | Not measured | 1362 | Only GPT-5.4 Nano measured |
| LMArena Text: creative writing | Not measured | 1315 | Only GPT-5.4 Nano measured |
| LMArena Vision | Not measured | 1196 | Only GPT-5.4 Nano measured |
| LMArena Text: English | Not measured | 1381 | Only GPT-5.4 Nano measured |
| LMArena Text: non-English | Not measured | 1359 | Only GPT-5.4 Nano measured |
| LMArena Text: Chinese | Not measured | 1391 | Only GPT-5.4 Nano measured |
| LMArena Text: German | Not measured | 1368 | Only GPT-5.4 Nano measured |
| LMArena Text: Japanese | Not measured | 1343 | Only GPT-5.4 Nano measured |
| LMArena Text: Korean | Not measured | 1321 | Only GPT-5.4 Nano measured |
| LMArena Text: French | Not measured | 1395 | Only GPT-5.4 Nano measured |
| LMArena Text: Spanish | Not measured | 1371 | Only GPT-5.4 Nano measured |
| LMArena Text: Russian | Not measured | 1363 | Only GPT-5.4 Nano measured |
| LMArena Text: Polish | Not measured | 1366 | Only GPT-5.4 Nano measured |
| LMArena Text: multi-turn | Not measured | 1382 | Only GPT-5.4 Nano measured |
| LMArena Text: longer queries | Not measured | 1365 | Only GPT-5.4 Nano measured |
| LMArena Text: writing and language | Not measured | 1338 | Only GPT-5.4 Nano measured |
| LMArena Text: entertainment and media | Not measured | 1325 | Only GPT-5.4 Nano measured |
| LMArena Vision: English | Not measured | 1202 | Only GPT-5.4 Nano measured |
| LMArena Vision: Chinese | Not measured | 1232 | Only GPT-5.4 Nano measured |
| LMArena Vision: reading text | Not measured | 1204 | Only GPT-5.4 Nano measured |
| LMArena Vision: humor | Not measured | 1161 | Only GPT-5.4 Nano measured |
| LMArena Vision: recognizing things | Not measured | 1141 | Only GPT-5.4 Nano measured |
| LMArena Vision: creative writing | Not measured | 1159 | Only GPT-5.4 Nano measured |
| Neither model has a result on the other 15 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Codestral 2508 | GPT-5.4 Nano | 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
Codestral 2508 costs less
Repository review
50K input + 3K output tokens
GPT-5.4 Nano costs less
Agent loop
200K input + 10K output tokens
GPT-5.4 Nano 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 Codestral 2508 | OpenAI GPT-5.4 Nano |
|---|---|---|
| Maker | Mistral AI | OpenAI |
| Weights | No public weights listed | No public weights listed |
| Released | August 1, 2025 | March 17, 2026 |
| Context window | 256K | 400K |
| Input $/Mtok | $0.30 | $0.20 |
| Output $/Mtok | $0.90 | $1.25 |
| Coding index | Not benchmarked | 44.8 |
| Independent measurements behind it | 0 | 2 |
| Cost per solved task | Not computable | $0.14 |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Codestral 2508 better than GPT-5.4 Nano for coding?
Only GPT-5.4 Nano has a coding score, 44.8 on the Vibe Coding Index; Codestral 2508 has none, so the two cannot be ranked on coding. Codestral 2508 has the edge where output tokens cost 1.4x less.
Which is cheaper, Codestral 2508 or GPT-5.4 Nano?
Codestral 2508 costs $0.30 per million input tokens and $0.90 per million output tokens; GPT-5.4 Nano costs $0.20 and $1.25. Cost per solved task is known for GPT-5.4 Nano only.
Which has the bigger context window, Codestral 2508 or GPT-5.4 Nano?
GPT-5.4 Nano, with 400K tokens against 256K.
How many benchmarks is the Codestral 2508 vs GPT-5.4 Nano comparison based on?
1 benchmarks have a result for both models as of October 21, 2026. Codestral 2508's coding index rests on 0 independent benchmarks and GPT-5.4 Nano's on 2. Where only one model has a result, the page says so instead of naming a winner.