Model comparison
Codestral vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.6 on the Noometry Index.
Last verified . 5 shared benchmarks.
Mistral AI
DeepSeekSummary
- They share 5 benchmarks with published results for both. Codestral scores higher in 0 categories and DeepSeek-V2.5 (Sep 2024) in 2 categories; 2 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek-V2.5 (Sep 2024) leads 25.6 to 19.8.
- The biggest single-benchmark swing is BigCodeBench Instruct: 41.8% for Codestral and 48.6% for DeepSeek-V2.5 (Sep 2024).
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| Codestral | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | Mistral AI | DeepSeek |
| Noometry Index | 30.6 | 37.6 |
| Released | 2024-05-29 | 2024-09-06 |
| Weights | Proprietary | Open |
| Context window | 256K | — |
| Max output | 8K | — |
| Input $ / M tokens | $0.30 | — |
| Output $ / M tokens | $0.90 | — |
| Results tracked | 7 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2.5 (Sep 2024) leads
Codestral: 27.3 (#321), DeepSeek-V2.5 (Sep 2024): 31.7 (#281)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| Aider Polyglot | 11.1% | 17.8% |
| BigCodeBench Instruct | 41.8% | 48.6% |
| BigCodeBench Complete | 52.5% | 53.2% |
| HumanEval+ | 73.8% | 83.5% |
| MBPP+ | 61.9% | 74.1% |
| LMArena Coding | — | 1309 |
| ALE-Bench | 137.78 | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Codestral: 19.8 (#251), DeepSeek-V2.5 (Sep 2024): 25.6 (#145)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| Kagi LLM Benchmark | 32.5% | — |
| LMArena Hard Prompts | — | 1289 |
Math Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 35.9 (#177)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | — | 1288 |
Knowledge Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 34.8 (#193)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | — | 1266 |
Multilingual Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 42.5 (#193)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | — | 1273 |
| LMArena Chinese | — | 1318 |
| LMArena French | — | 1289 |
| LMArena German | — | 1258 |
| LMArena Japanese | — | 1228 |
| LMArena Korean | — | 1209 |
| LMArena Russian | — | 1289 |
| LMArena Spanish | — | 1248 |
Instruction Following Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 67.5 (#194)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | — | 1280 |
Long Context Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 39.5 (#174)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | — | 1301 |
Writing & Preference Not comparable
Codestral: —, DeepSeek-V2.5 (Sep 2024): 49.8 (#187)
| Benchmark | Codestral | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | — | 1294 |
| LMArena Creative Writing | — | 1285 |
| LMArena Multi-Turn | — | 1297 |
Frequently asked questions
Is Codestral better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 30.6 on the Noometry Index.
Is Codestral or DeepSeek-V2.5 (Sep 2024) better for coding?
DeepSeek-V2.5 (Sep 2024) scores higher on coding benchmarks: 31.7 versus 27.3 in the Noometry coding category.
How many benchmarks do Codestral and DeepSeek-V2.5 (Sep 2024) share?
5 benchmarks have published results for both models. Codestral has 7 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.