Model comparison
Codestral vs GPT-5.4 nano
GPT-5.4 nano is the stronger model overall, scoring 41.9 to 30.1 on the Noometry Index.
Last verified . 2 shared benchmarks.
Mistral AISummary
- They share 2 benchmarks with published results for both. Codestral scores higher in 0 categories and GPT-5.4 nano in 2 categories; 2 gaps are clear of the uncertainty.
- The widest gap is in coding, where GPT-5.4 nano leads 43.1 to 26.9.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 32.5% for Codestral and 39.7% for GPT-5.4 nano.
- Both cost about the same: $0.30 input and $0.90 output per million tokens.
- GPT-5.4 nano accepts more context: 400K tokens versus 256K.
- Codestral has downloadable open weights; the other is API-only.
Side by side
| Codestral | GPT-5.4 nano | |
|---|---|---|
| Provider | Mistral AI | OpenAI |
| Noometry Index | 30.1 | 41.9 |
| Released | 2024-05-29 | 2026-03-17 |
| Weights | Open | Proprietary |
| Context window | 256K | 400K |
| Max output | 4K | 128K |
| Input $ / M tokens | $0.30 | $0.20 |
| Output $ / M tokens | $0.90 | $1.25 |
| Results tracked | 7 | 40 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding GPT-5.4 nano leads
Codestral: 26.9 (#316), GPT-5.4 nano: 43.1 (#82)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| ALE-Bench | 137.78 | 1,005 |
| Aider Polyglot | 11.1% | — |
| SciCode | — | 46.9% |
| WeirdML | — | 49.2% |
| BigCodeBench Instruct | 41.8% | — |
| LMArena Coding | — | 1404 |
| BigCodeBench Complete | 52.5% | — |
| HumanEval+ | 73.8% | — |
| MBPP+ | 61.9% | — |
Reasoning GPT-5.4 nano leads
Codestral: 19.1 (#244), GPT-5.4 nano: 23.8 (#159)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| Kagi LLM Benchmark | 32.5% | 39.7% |
| ARC-AGI-2 | — | 5.7% |
| ARC-AGI-1 | — | 51.5% |
| CritPt | — | 9.3% |
| Chess Puzzles | — | 30% |
| LMArena Hard Prompts | — | 1380 |
| Mystery Game Puzzles | — | 9% |
| DTBench | — | 80.3% |
| LMCA | — | 36.9% |
| Epoch Capabilities Index | — | 145.78 |
| ForecastBench | — | 57.3 |
Math Not comparable
Codestral: —, GPT-5.4 nano: 40.7 (#76)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| FrontierMath (Tiers 1-3) | — | 44.9% |
| FrontierMath Tier 4 | — | 12.2% |
| OTIS Mock AIME 2024-2025 | — | 87.8% |
| ProofBench | — | 5% |
| LMArena Math | — | 1406 |
| FrontierMath (Feb 2025 set) | — | 25.9% |
| FrontierMath Tier 4 (v1) | — | 6.3% |
Knowledge Not comparable
Codestral: —, GPT-5.4 nano: 41.8 (#94)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| GPQA Diamond | — | 78.5% |
| SimpleQA Verified | — | 11.7% |
| Vectara Hallucination Rate | — | 3.1% |
| LMArena Expert | — | 1397 |
Multimodal Not comparable
Codestral: —, GPT-5.4 nano: 36.0 (#73)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| LMArena Vision | — | 1196 |
Multilingual Not comparable
Codestral: —, GPT-5.4 nano: 48.8 (#136)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| LMArena Non-English | — | 1359 |
| LMArena Chinese | — | 1391 |
| LMArena French | — | 1396 |
| LMArena German | — | 1368 |
| LMArena Japanese | — | 1342 |
| LMArena Korean | — | 1321 |
| LMArena Russian | — | 1363 |
| LMArena Spanish | — | 1368 |
Instruction Following Not comparable
Codestral: —, GPT-5.4 nano: 72.1 (#139)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| LMArena Instruction Following | — | 1362 |
Long Context Not comparable
Codestral: —, GPT-5.4 nano: 41.6 (#131)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| LMArena Longer Query | — | 1365 |
Writing & Preference Not comparable
Codestral: —, GPT-5.4 nano: 55.9 (#137)
| Benchmark | Codestral | GPT-5.4 nano |
|---|---|---|
| LMArena Text | — | 1372 |
| LMArena Creative Writing | — | 1313 |
| LMArena Multi-Turn | — | 1382 |
Frequently asked questions
Is Codestral better than GPT-5.4 nano?
GPT-5.4 nano is the stronger model overall, scoring 41.9 to 30.1 on the Noometry Index.
Which is cheaper, Codestral or GPT-5.4 nano?
Codestral is cheaper. It lists at $0.30 per million input tokens and $0.90 per million output tokens; GPT-5.4 nano lists at $0.20 and $1.25.
Is Codestral or GPT-5.4 nano better for coding?
GPT-5.4 nano scores higher on coding benchmarks: 43.1 versus 26.9 in the Noometry coding category.
Which has the bigger context window?
GPT-5.4 nano does, with 400K tokens against 256K.
How many benchmarks do Codestral and GPT-5.4 nano share?
2 benchmarks have published results for both models. Codestral has 7 scored results on Noometry and GPT-5.4 nano has 40.