Model comparison
Codestral vs DeepSeek-V2 (MoE-236B, May 2024)
Codestral has enough public results to be ranked (#290); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Last verified . 2 shared benchmarks.
Mistral AI
DeepSeekSummary
- They share 2 benchmarks with published results for both. Codestral scores higher in 0 categories and DeepSeek-V2 (MoE-236B, May 2024) in 1 category; one gap is clear of the uncertainty.
- The widest gap is in coding, where DeepSeek-V2 (MoE-236B, May 2024) leads 40.4 to 27.3.
- The biggest single-benchmark swing is BigCodeBench Instruct: 41.8% for Codestral and 48.9% for DeepSeek-V2 (MoE-236B, May 2024).
- DeepSeek-V2 (MoE-236B, May 2024) has downloadable open weights; the other is API-only.
Side by side
| Codestral | DeepSeek-V2 (MoE-236B, May 2024) | |
|---|---|---|
| Provider | Mistral AI | DeepSeek |
| Noometry Index | 30.6 | 40.3 |
| Released | 2024-05-29 | 2024-05-07 |
| Weights | Proprietary | Open |
| Context window | 256K | — |
| Max output | 8K | — |
| Input $ / M tokens | $0.30 | — |
| Output $ / M tokens | $0.90 | — |
| Results tracked | 7 | 10 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2 (MoE-236B, May 2024) leads
Codestral: 27.3 (#321), DeepSeek-V2 (MoE-236B, May 2024): 40.4 (#139)
| Benchmark | Codestral | DeepSeek-V2 (MoE-236B, May 2024) |
|---|---|---|
| BigCodeBench Instruct | 41.8% | 48.9% |
| BigCodeBench Complete | 52.5% | 59.4% |
| Aider Polyglot | 11.1% | — |
| ALE-Bench | 137.78 | — |
| HumanEval+ | 73.8% | — |
| MBPP+ | 61.9% | — |
Reasoning Not comparable
Codestral: 19.8 (#251), DeepSeek-V2 (MoE-236B, May 2024): —
| Benchmark | Codestral | DeepSeek-V2 (MoE-236B, May 2024) |
|---|---|---|
| Kagi LLM Benchmark | 32.5% | — |
| BIG-Bench Hard | — | 78.8% |
| Epoch Capabilities Index | — | 124.77 |
| HellaSwag | — | 87.1% |
| PIQA | — | 83.9% |
| WinoGrande | — | 86.3% |
Knowledge Not comparable
Codestral: —, DeepSeek-V2 (MoE-236B, May 2024): —
| Benchmark | Codestral | DeepSeek-V2 (MoE-236B, May 2024) |
|---|---|---|
| ARC (AI2) Challenge | — | 92.2% |
| MMLU | — | 78.4% |
| TriviaQA | — | 80% |
Frequently asked questions
Is Codestral better than DeepSeek-V2 (MoE-236B, May 2024)?
Codestral has enough public results to be ranked (#290); DeepSeek-V2 (MoE-236B, May 2024) does not yet, so treat this comparison as directional.
Is Codestral or DeepSeek-V2 (MoE-236B, May 2024) better for coding?
DeepSeek-V2 (MoE-236B, May 2024) scores higher on coding benchmarks: 40.4 versus 27.3 in the Noometry coding category.
How many benchmarks do Codestral and DeepSeek-V2 (MoE-236B, May 2024) share?
2 benchmarks have published results for both models. Codestral has 7 scored results on Noometry and DeepSeek-V2 (MoE-236B, May 2024) has 10.