Model comparison
GPT-4 vs Mistral Small
Mistral Small is the stronger model overall, scoring 33.4 to 29.1 on the Noometry Index.
Last verified . 25 shared benchmarks.
Mistral AISummary
- They share 25 benchmarks with published results for both. GPT-4 scores higher in 0 categories and Mistral Small in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Mistral Small leads 52.5 to 34.9.
- The biggest single-benchmark swing is MATH Level 5: 23% for GPT-4 and 46.8% for Mistral Small.
- Mistral Small is cheaper at $0.15 / $0.60 per million input/output tokens, against $30 / $60 for GPT-4.
- Mistral Small accepts more context: 262K tokens versus 8K.
- Mistral Small has downloadable open weights; the other is API-only.
Side by side
| GPT-4 | Mistral Small | |
|---|---|---|
| Provider | OpenAI | Mistral AI |
| Noometry Index | 29.1 | 33.4 |
| Released | 2023-03-14 | 2024-02-26 |
| Weights | Proprietary | Open |
| Context window | 8K | 262K |
| Max output | 8K | 256K |
| Input $ / M tokens | $30 | $0.15 |
| Output $ / M tokens | $60 | $0.60 |
| Results tracked | 38 | 39 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Mistral Small leads
GPT-4: 31.6 (#283), Mistral Small: 34.0 (#247)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| BigCodeBench Instruct | 46% | 36.1% |
| LMArena Coding | 1254 | 1362 |
| BigCodeBench Complete | 57.2% | 46.6% |
| SciCode | — | 26.5% |
| WeirdML | 12.4% | — |
| LiveBench Coding | — | 36.2% |
| ALE-Bench | — | 497.62 |
| HumanEval+ | 79.3% | — |
Agentic & Tool Use Not comparable
GPT-4: —, Mistral Small: 28.1 (#93)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| Berkeley Function Calling Leaderboard | — | 37.1% |
| METR Time Horizons | 36.1% | — |
Reasoning Mistral Small leads
GPT-4: 17.8 (#289), Mistral Small: 19.8 (#250)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Hard Prompts | 1241 | 1335 |
| DTBench | 62.7% | 70.9% |
| LMCA | 17.1% | 20.6% |
| Kagi LLM Benchmark | — | 37.8% |
| CritPt | — | 0% |
| Chess Puzzles | 4% | — |
| LiveBench Reasoning | — | 44.8% |
| Mystery Game Puzzles | 12% | — |
| LiveBench Data Analysis | — | 53.7% |
| BIG-Bench Hard | 75.1% | — |
| Epoch Capabilities Index | 125.89 | — |
| ForecastBench | 57.8 | — |
| HellaSwag | 95.3% | — |
| LiveBench | — | 44% |
| WinoGrande | 87.5% | — |
Math Mistral Small leads
GPT-4: 10.8 (#309), Mistral Small: 16.4 (#293)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.1% | 5.8% |
| LMArena Math | 1269 | 1341 |
| MATH Level 5 | 23% | 46.8% |
| LiveBench Math | — | 39.9% |
| GSM8K | 92% | — |
Knowledge Mistral Small leads
GPT-4: 18.4 (#282), Mistral Small: 31.0 (#222)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| GPQA Diamond | 35.7% | 47.5% |
| LMArena Expert | 1211 | 1291 |
| MMLU | 86.4% | 68.7% |
| Vectara Hallucination Rate | — | 5.1% |
| TriviaQA | 84.8% | — |
Multimodal Not comparable
GPT-4: —, Mistral Small: 33.5 (#96)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Vision | — | 1142 |
Multilingual Mistral Small leads
GPT-4: 40.6 (#215), Mistral Small: 45.5 (#169)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Non-English | 1246 | 1315 |
| LMArena Chinese | 1242 | 1340 |
| LMArena French | 1283 | 1337 |
| LMArena German | 1251 | 1340 |
| LMArena Japanese | 1209 | 1275 |
| LMArena Korean | 1184 | 1259 |
| LMArena Russian | 1251 | 1324 |
| LMArena Spanish | 1261 | 1346 |
Instruction Following Mistral Small leads
GPT-4: 65.3 (#222), Mistral Small: 66.4 (#209)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Instruction Following | 1241 | 1310 |
| LiveBench Instruction Following | — | 63.7% |
Long Context Mistral Small leads
GPT-4: 37.7 (#212), Mistral Small: 40.4 (#156)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Longer Query | 1244 | 1327 |
Writing & Preference Mistral Small leads
GPT-4: 34.9 (#268), Mistral Small: 52.5 (#171)
| Benchmark | GPT-4 | Mistral Small |
|---|---|---|
| LMArena Text | 1263 | 1338 |
| LMArena Creative Writing | 1244 | 1305 |
| LMArena Multi-Turn | 1257 | 1344 |
| EQ-Bench Creative Writing | 752 | — |
| LiveBench Language | — | 30.5% |
Frequently asked questions
Is GPT-4 better than Mistral Small?
Mistral Small is the stronger model overall, scoring 33.4 to 29.1 on the Noometry Index.
Which is cheaper, GPT-4 or Mistral Small?
Mistral Small is cheaper. It lists at $0.15 per million input tokens and $0.60 per million output tokens; GPT-4 lists at $30 and $60.
Is GPT-4 or Mistral Small better for coding?
Mistral Small scores higher on coding benchmarks: 34.0 versus 31.6 in the Noometry coding category.
Which has the bigger context window?
Mistral Small does, with 262K tokens against 8K.
How many benchmarks do GPT-4 and Mistral Small share?
25 benchmarks have published results for both models. GPT-4 has 38 scored results on Noometry and Mistral Small has 39.