BlogLLM Benchmark Scores as Published, Side by Side
DeepSeek-V3: Benchmark Scores as Published
Every result the catalog holds for DeepSeek-V3: 23 tests in 9 areas, from 3 pages. 21 printed by DeepSeek itself, the rest by the company named beside each number. Each number is as printed; nothing is ranked or rescored.
Developer DeepSeekReleased 2024-12-26Context window 128KInput TextOutput TextWeights Open (Model License)Data version 2026103001, as of 2026-10-30
Agentic Coding
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| SWE-bench Verified | 42% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Science and Math
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| GPQA Diamond | 59.1% | DeepSeek, DeepSeek-V3-0324 model cardchecked 2026-10-15; printed on 2 more pages | - |
| AIME 2025 | 28.8% | Mistral AI, Announcing Magistralchecked 2026-10-15 | Attempt budget: pass@1 |
| AIME 2024 | 39.6% | DeepSeek, DeepSeek-V3-0324 model cardchecked 2026-10-15; printed on 2 more pages | - |
| MATH-500 | 90.2% | DeepSeek, DeepSeek-V3-0324 model cardchecked 2026-10-15; printed on 1 more page | - |
| CNMO 2024 | 43.2% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
General Knowledge and Reasoning
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| MMLU-Pro | 75.9% | DeepSeek, DeepSeek-V3-0324 model cardchecked 2026-10-15; printed on 1 more page | - |
| MMLU-Redux | 89.1% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| C-Eval | 86.5% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| MMLU | 88.5% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| SimpleQA | 24.9% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | Tools: none |
| DROP (F1) | 91.6% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| Chinese SimpleQA | 68% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Code Generation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Codeforces rating | 1134 | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| LiveCodeBench (version not stated) | 39.2% | DeepSeek, DeepSeek-V3-0324 model cardchecked 2026-10-15 | - |
| Aider polyglot | 49.6% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15; printed on 1 more page | - |
| Codeforces percentile | 58.7% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
| LiveCodeBench v5 | 36.2% | Mistral AI, Announcing Magistralchecked 2026-10-15 | - |
Web Search and Research
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| FRAMES | 73.3% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Instruction Following
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| IFEval (prompt-level strict) | 86.1% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Multilingual
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| CLUEWSC | 90.9% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Writing and Conversation
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| AlpacaEval 2.0 (length-controlled win rate) | 70% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
Preference and Composite Rankings
| Test | Value | Printed by | Setup as printed |
|---|---|---|---|
| Arena-Hard win rate (GPT-4-1106 judge) | 85.5% | DeepSeek, DeepSeek-R1 model cardchecked 2026-10-15 | - |
More from DeepSeek
- DeepSeek-V4.1-Flash-1017
- DeepSeek-V4.1-Flash
- DeepSeek-V4-Flash (Preview)
- DeepSeek-V4-Flash-Base
- DeepSeek-V4-Pro (Preview)
- DeepSeek-V4-Pro-Base
- DeepSeek-V3.2
- DeepSeek-V3.2-Base
- DeepSeek-V3.2-Speciale
- DeepSeek-V3.2-Exp
- DeepSeek-V3.1-Terminus
- DeepSeek-V3.1
- DeepSeek-R1-0528
- DeepSeek-R1-0528-Qwen3-8B
- DeepSeek-V3-0324
- DeepSeek-R1
- DeepSeek-R1-Distill-Llama-70B
- DeepSeek-R1-Distill-Llama-8B
- DeepSeek-R1-Distill-Qwen-1.5B
- DeepSeek-R1-Distill-Qwen-14B
- DeepSeek-R1-Distill-Qwen-32B
- DeepSeek-R1-Distill-Qwen-7B
- DeepSeek-V4-Flash-0731
- DeepSeek-V4-Pro-0813
- DeepSeek-V4-Vision-Exp
Every number here is read from the same catalog file as the benchmark table, data version 2026103001. Where one page printed the model under several setups, each value is listed. The catalog is public in armanobosyan/sugra-llm-benchmarks; a wrong number can be reported from the table, with Report an error under the value, or as an issue in that repository.