DeepSeek V3 vs Mistral Large 2: Benchmark Comparison
Detailed comparison of DeepSeek V3 and Mistral Large 2 covering benchmarks, pricing, context window, and compliance.
Key Specifications
| Specification | DeepSeek V3 | Mistral Large 2 |
|---|---|---|
| Vendor | deepseek | mistral |
| Version | v3 | large-2 |
| Release Date | 2024-12-26 | 2024-07-24 |
| Context Window | 64000 tokens | 128000 tokens |
| Input Modalities | text | text |
| Output Modalities | text | text |
| License | DeepSeek License | Mistral Research License |
| SOC2 | ✗ | ✓ |
| HIPAA | ✗ | ✗ |
| GDPR | ✗ | ✓ |
| ISO 27001 | ✗ | ✓ |
Benchmark Results
| Benchmark | DeepSeek V3 | Mistral Large 2 | Winner |
|---|---|---|---|
| BBH | 84.9 | 81 | DeepSeek V3 |
| GSM8K | 89.3 | 93 | Mistral Large 2 |
| HUMANEVAL | 82.6 | 92 | Mistral Large 2 |
| MATH | 61.6 | 71 | Mistral Large 2 |
| MMLU | 88.5 | 84 | DeepSeek V3 |
Pricing Comparison
| Tier (per Mtok) | DeepSeek V3 | Mistral Large 2 |
|---|---|---|
| Input | $0.27 | $2 |
| Output | $1.1 | $6 |
| Cache Read | $0.07 | $0 |
| Cache Write | $0.27 | $0 |
DeepSeek V3 対 Mistral Large 2
モデル概要
DeepSeek V3 and Mistral Large 2 are both notable options in the AI model market. This page compares their benchmarks, pricing, and compliance.
主要仕様
| ベンダー | リリース日 | コンテキストウィンドウ | ライセンス |
|---|---|---|---|
| Deepseek / Mistral | 2024-12-26 / 2024-07-24 | 64K / 128K | DeepSeek License / Mistral Research License |
ベンチマークパフォーマンス
| ベンチマーク | DeepSeek V3 | Mistral Large 2 | 勝者 |
|---|---|---|---|
| BBH (BIG-Bench Hard) | 84.9 | 81.0 | A |
| GSM8K (Grade School Math 8K) | 89.3 | 93.0 | B |
| HumanEval | 82.6 | 92.0 | B |
| MATH | 61.6 | 71.0 | B |
| MMLU (Massive Multitask Language Understanding) | 88.5 | 84.0 | A |
料金比較
| 入力 | 出力 | キャッシュ読み取り | キャッシュ書き込み |
|---|---|---|---|
| — / — | — / — | — / — | — / — |
100万トークンあたり — A / B
強み & 弱み
DeepSeek V3
- ✅ MMLU score 88.5, strong knowledge reasoning.
- ✅ HumanEval 82.6, excellent code generation.
- ✅ GSM8K 89.3, robust math reasoning.
- ✅ 采用 MoE 混合专家架构。
- ⚠️ 闭源专有模型，不支持自托管。
Mistral Large 2
- ✅ MMLU score 84.0, strong knowledge reasoning.
- ✅ HumanEval 92.0, excellent code generation.
- ✅ GSM8K 93.0, robust math reasoning.
- ⚠️ 闭源专有模型，不支持自托管。
編集者コメント
DeepSeek V3 and Mistral Large 2 each have their strengths. Choose based on workload (code, long context, vision), referencing the tables above.
FAQ
Which model is better for coding tasks?
Refer to the HumanEval benchmark table; the model with a higher score is better suited for coding tasks.
Which model is cheaper?
Refer to the pricing comparison table above; the model with lower input/output prices is more cost-effective.
Which has a longer context window?
Refer to the key specifications table; the model with a larger context window is better for long documents.
参考文献
Editor's Take
See Editor's Take section.