Model comparison
Codestral vs DeepSeek-R1
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 30.6 on the Noometry Index. Codestral costs 2.0× less per token, which makes it the better buy when DeepSeek-R1's lead doesn't matter for your workload.
Last verified . 3 shared benchmarks.
Mistral AI
DeepSeekSummary
- They share 3 benchmarks with published results for both. Codestral scores higher in 1 category and DeepSeek-R1 in 1 category; 2 gaps are clear of the uncertainty.
- The widest gap is in coding, where DeepSeek-R1 leads 46.3 to 27.3.
- The biggest single-benchmark swing is Aider Polyglot: 11.1% for Codestral and 71.4% for DeepSeek-R1.
- Codestral is cheaper at $0.30 / $0.90 per million input/output tokens, against $0.50 / $2.15 for DeepSeek-R1.
- Codestral accepts more context: 256K tokens versus 164K.
Side by side
| Codestral | DeepSeek-R1 | |
|---|---|---|
| Provider | Mistral AI | DeepSeek |
| Noometry Index | 30.6 | 42.3 |
| Released | 2024-05-29 | 2025-01-20 |
| Weights | Proprietary | Proprietary |
| Context window | 256K | 164K |
| Max output | 8K | 64K |
| Input $ / M tokens | $0.30 | $0.50 |
| Output $ / M tokens | $0.90 | $2.15 |
| Results tracked | 7 | 52 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-R1 leads
Codestral: 27.3 (#321), DeepSeek-R1: 46.3 (#68)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| Aider Polyglot | 11.1% | 71.4% |
| ALE-Bench | 137.78 | 804.12 |
| SciCode | — | 35.7% |
| WeirdML | — | 41.6% |
| BigCodeBench Instruct | 41.8% | — |
| LiveBench Coding | — | 66.7% |
| LMArena Coding | — | 1427 |
| BigCodeBench Complete | 52.5% | — |
| AlgoTune | — | 1.7 |
| HumanEval+ | 73.8% | — |
| MBPP+ | 61.9% | — |
Agentic & Tool Use Not comparable
Codestral: —, DeepSeek-R1: 30.7 (#75)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| DeepResearch Bench | — | 35.1% |
| BALROG | — | 34.9% |
| METR Time Horizons | — | 53.8% |
Reasoning Codestral leads
Codestral: 19.8 (#251), DeepSeek-R1: 18.6 (#278)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| Kagi LLM Benchmark | 32.5% | 69.4% |
| ARC-AGI-2 | — | 1.3% |
| SimpleBench | — | 40.8% |
| ARC-AGI-1 | — | 21.2% |
| CritPt | — | 1.1% |
| LiveBench Reasoning | — | 83.2% |
| LMArena Hard Prompts | — | 1416 |
| LiveBench Data Analysis | — | 69.8% |
| Epoch Capabilities Index | — | 141.29 |
| ForecastBench | — | 60 |
| LiveBench | — | 71.6% |
Math Not comparable
Codestral: —, DeepSeek-R1: 43.8 (#79)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| OTIS Mock AIME 2024-2025 | — | 66.4% |
| Omni-MATH | — | 42.4% |
| LiveBench Math | — | 80.7% |
| LMArena Math | — | 1400 |
| MATH Level 5 | — | 96.6% |
Knowledge Not comparable
Codestral: —, DeepSeek-R1: 44.5 (#87)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| GPQA Diamond | — | 76.3% |
| MMLU-Pro | — | 79.3% |
| Confabulations | — | 12.7% |
| Vectara Hallucination Rate | — | 11.3% |
| GPQA (HELM) | — | 66.6% |
| LMArena Expert | — | 1394 |
Multilingual Not comparable
Codestral: —, DeepSeek-R1: 52.4 (#85)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| LMArena Non-English | — | 1412 |
| LMArena Chinese | — | 1442 |
| LMArena French | — | 1417 |
| LMArena German | — | 1404 |
| LMArena Japanese | — | 1391 |
| LMArena Korean | — | 1360 |
| LMArena Russian | — | 1423 |
| LMArena Spanish | — | 1411 |
Instruction Following Not comparable
Codestral: —, DeepSeek-R1: 72.0 (#143)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| LiveBench Instruction Following | — | 80.5% |
| IFEval | — | 78.4% |
| LMArena Instruction Following | — | 1382 |
Long Context Not comparable
Codestral: —, DeepSeek-R1: 45.4 (#36)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| Fiction.LiveBench | — | 75% |
| LMArena Longer Query | — | 1391 |
Writing & Preference Not comparable
Codestral: —, DeepSeek-R1: 61.4 (#88)
| Benchmark | Codestral | DeepSeek-R1 |
|---|---|---|
| LMArena Text | — | 1428 |
| LMArena Creative Writing | — | 1405 |
| Short-Story Creative Writing | — | 83% |
| EQ-Bench Creative Writing | — | 1500 |
| WildBench | — | 82.8% |
| LMArena Multi-Turn | — | 1405 |
| LiveBench Language | — | 48.5% |
Frequently asked questions
Is Codestral better than DeepSeek-R1?
DeepSeek-R1 is the stronger model overall, scoring 42.3 to 30.6 on the Noometry Index. Codestral costs 2.0× less per token, which makes it the better buy when DeepSeek-R1's lead doesn't matter for your workload.
Which is cheaper, Codestral or DeepSeek-R1?
Codestral is cheaper. It lists at $0.30 per million input tokens and $0.90 per million output tokens; DeepSeek-R1 lists at $0.50 and $2.15.
Is Codestral or DeepSeek-R1 better for coding?
DeepSeek-R1 scores higher on coding benchmarks: 46.3 versus 27.3 in the Noometry coding category.
Which has the bigger context window?
Codestral does, with 256K tokens against 164K.
How many benchmarks do Codestral and DeepSeek-R1 share?
3 benchmarks have published results for both models. Codestral has 7 scored results on Noometry and DeepSeek-R1 has 52.