Model comparison
GPT-3.5-turbo vs Mixtral 8x22B
Mixtral 8x22B is the stronger model overall, scoring 27.1 to 23.2 on the Noometry Index. GPT-3.5-turbo costs 4.0× less per token, which makes it the better buy when Mixtral 8x22B's lead doesn't matter for your workload.
Last verified . 28 shared benchmarks.
Mistral AISummary
- They share 28 benchmarks with published results for both. GPT-3.5-turbo scores higher in 1 category and Mixtral 8x22B in 7 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in math, where Mixtral 8x22B leads 22.9 to 6.3.
- The biggest single-benchmark swing is MATH Level 5: 15.9% for GPT-3.5-turbo and 24.2% for Mixtral 8x22B.
- GPT-3.5-turbo is cheaper at $0.50 / $1.50 per million input/output tokens, against $2 / $6 for Mixtral 8x22B.
- Mixtral 8x22B accepts more context: 64K tokens versus 16K.
- Mixtral 8x22B has downloadable open weights; the other is API-only.
Side by side
| GPT-3.5-turbo | Mixtral 8x22B | |
|---|---|---|
| Provider | OpenAI | Mistral AI |
| Noometry Index | 23.2 | 27.1 |
| Released | 2023-03-01 | 2024-04-17 |
| Weights | Proprietary | Open |
| Context window | 16K | 64K |
| Max output | 4K | 64K |
| Input $ / M tokens | $0.50 | $2 |
| Output $ / M tokens | $1.50 | $6 |
| Results tracked | 44 | 34 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
GPT-3.5-turbo: 23.9 (#331), Mixtral 8x22B: 24.2 (#329)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| WeirdML | 3.5% | 3.2% |
| BigCodeBench Instruct | 39.1% | 40.6% |
| LMArena Coding | 1136 | 1166 |
| BigCodeBench Complete | 50.6% | 50.2% |
| HumanEval+ | 70.7% | 72% |
| MBPP+ | 69.7% | 64.3% |
Agentic & Tool Use Not comparable
GPT-3.5-turbo: —, Mixtral 8x22B: 23.1 (#127)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| Cybench | — | 7.5% |
| METR Time Horizons | 21.5% | — |
Reasoning Mixtral 8x22B leads
GPT-3.5-turbo: 13.8 (#332), Mixtral 8x22B: 19.9 (#248)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Hard Prompts | 1108 | 1150 |
| DTBench | 48.5% | 55.1% |
| Epoch Capabilities Index | 118.55 | 122.03 |
| ForecastBench | 50.4 | 56.3 |
| Chess Puzzles | 0% | — |
| Mystery Game Puzzles | 3% | — |
| LMCA | 9.7% | — |
| Adversarial NLI | 58.1% | — |
| BIG-Bench Hard | 61.6% | — |
| CommonsenseQA 2.0 | 57% | — |
| WinoGrande | 81.6% | — |
Math Mixtral 8x22B leads
GPT-3.5-turbo: 6.3 (#327), Mixtral 8x22B: 22.9 (#275)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Math | 1142 | 1184 |
| MATH Level 5 | 15.9% | 24.2% |
| FrontierMath (Tiers 1-3) | 0% | — |
| OTIS Mock AIME 2024-2025 | 2.2% | — |
| Omni-MATH | — | 16.3% |
| GSM8K | 57.8% | — |
Knowledge Mixtral 8x22B leads
GPT-3.5-turbo: 10.0 (#303), Mixtral 8x22B: 15.1 (#293)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| GPQA Diamond | 28% | 34.1% |
| LMArena Expert | 1070 | 1113 |
| MMLU | 71.4% | 77.8% |
| MMLU-Pro | — | 46% |
| GPQA (HELM) | — | 33.4% |
| ARC (AI2) Challenge | 87.4% | — |
| BoolQ | 87% | — |
| OpenBookQA | 86% | — |
| TriviaQA | 85.8% | — |
Multilingual Mixtral 8x22B leads
GPT-3.5-turbo: 31.5 (#258), Mixtral 8x22B: 32.8 (#255)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Non-English | 1108 | 1128 |
| LMArena Chinese | 1075 | 1116 |
| LMArena French | 1118 | 1166 |
| LMArena German | 1090 | 1141 |
| LMArena Japanese | 1043 | 1037 |
| LMArena Korean | 1019 | 1057 |
| LMArena Russian | 1123 | 1158 |
| LMArena Spanish | 1121 | 1151 |
Instruction Following Too close to call
GPT-3.5-turbo: 57.9 (#262), Mixtral 8x22B: 57.7 (#266)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Instruction Following | 1119 | 1147 |
| IFEval | — | 72.4% |
Long Context Too close to call
GPT-3.5-turbo: 34.0 (#254), Mixtral 8x22B: 34.7 (#247)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Longer Query | 1121 | 1144 |
Writing & Preference Mixtral 8x22B leads
GPT-3.5-turbo: 25.3 (#305), Mixtral 8x22B: 36.9 (#262)
| Benchmark | GPT-3.5-turbo | Mixtral 8x22B |
|---|---|---|
| LMArena Text | 1125 | 1162 |
| LMArena Creative Writing | 1092 | 1141 |
| LMArena Multi-Turn | 1117 | 1130 |
| EQ-Bench Creative Writing | 451 | — |
| WildBench | — | 71.1% |
Frequently asked questions
Is GPT-3.5-turbo better than Mixtral 8x22B?
Mixtral 8x22B is the stronger model overall, scoring 27.1 to 23.2 on the Noometry Index. GPT-3.5-turbo costs 4.0× less per token, which makes it the better buy when Mixtral 8x22B's lead doesn't matter for your workload.
Which is cheaper, GPT-3.5-turbo or Mixtral 8x22B?
GPT-3.5-turbo is cheaper. It lists at $0.50 per million input tokens and $1.50 per million output tokens; Mixtral 8x22B lists at $2 and $6.
Is GPT-3.5-turbo or Mixtral 8x22B better for coding?
They score almost the same on coding (23.9 vs 24.2); test both on your own repository before choosing.
Which has the bigger context window?
Mixtral 8x22B does, with 64K tokens against 16K.
How many benchmarks do GPT-3.5-turbo and Mixtral 8x22B share?
28 benchmarks have published results for both models. GPT-3.5-turbo has 44 scored results on Noometry and Mixtral 8x22B has 34.