DeepSeek V4 Pro
| Released | 2026-04-24 |
|---|---|
| Price per 1M | $1.65 in · $3.30 out |
| Overall score | 55.6 · rank 71 |
| Context window | 1,000,000 tokens |
| Provider | DeepSeek |
| Weights | Open weights |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 55.6 | 71 | 32 |
| Overall: Coding | 55.0 | 71 | 19 |
| Overall: Reasoning | 55.9 | 67 | 15 |
| Overall: Agentic work | 56.4 | 67 | 17 |
| Overall: Frontend | 53.1 | 70 | 7 |
| Coding | 54.4 | 80 | 5 |
| Agents | 56.9 | 62 | 5 |
| Reasoning | 58.0 | 65 | 7 |
| Math | 54.0 | 78 | 5 |
| Knowledge | 54.4 | 66 | 3 |
| Frontend | 52.4 | 64 | 2 |
| General | 56.9 | 48 | 5 |
| Value vs price | +5.7 | 69 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 52.3 ±3.7 | 9 |
| High | 55.9 ±3.1 | 9 |
| Max | 56.2 ±1.7 | 19 |
| Non-reasoning | 48.5 ±3.8 | 5 |
Evidence
Coding 54.4 ±2.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierCode | 17.6% | Epoch AI | 2026-11-11 |
| SciCode | 50.0% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 87.5% | Vals AI | 2026-09-01 |
| SWE-bench Verified (Epoch run) | 77.6% | Epoch AI | 2026-06-18 |
| ProgramBench | 0.0% | Vals AI | 2026-11-09 |
| SWE-bench Verified (Vals run) | 77.4% | Vals AI | 2026-09-01 |
Agents 56.9 ±2.5
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 4.0 (Vals run) | 11.1% | Vals AI | 2026-11-09 |
| Terminal-Bench 2.1 (Vals run) | 50.2% | Vals AI | 2026-10-01 |
| Tax Agent Bench | 25.2% | Vals AI | 2026-11-09 |
| SkillsBench | 51.3% | Vals AI | 2026-10-29 |
| Vending-Bench 2 | $3,285 | Epoch AI | 2026-11-11 |
Reasoning 58.0 ±4.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Extended NYT Connections | 91.3% | Lech Mazur | 2026-11-10 |
| CritPt | 12.9% | Epoch AI | 2026-11-11 |
| SimpleBench | 50.9% | SimpleBench | 2026-06-02 |
| GPQA Diamond | 90.9% | Epoch AI | 2026-08-06 |
| GPQA Diamond (Vals run) | 89.4% | Vals AI | 2026-09-01 |
| Kagi Reasoning Benchmark | 53.5% | Kagi | 2026-11-11 |
| Chess Puzzles | 19.0% | Epoch AI | 2026-10-13 |
| Mystery Game Puzzles | 17.0% | Epoch AI | 2026-08-27 |
Math 54.0 ±2.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierMath Tiers 1-3 | 45.3% | Epoch AI | 2026-06-17 |
| FrontierMath Tier 4 | 2.4% | Epoch AI | 2026-06-17 |
| OTIS Mock AIME 2024-2025 | 96.7% | Epoch AI | 2026-06-17 |
| ProofBench v1.1 | 16.0% | Vals AI | 2026-11-09 |
| MathArena Final-Answer Competitions | 76.6% | MathArena | 2026-11-11 |
| ProofBench | 16.0% | Epoch AI | 2026-11-11 |
Knowledge 54.4 ±3.6
| Benchmark | Result | Source | Date |
|---|---|---|---|
| MMLU-Pro (Vals run) | 87.2% | Vals AI | 2026-09-01 |
| SimpleQA Verified | 47.0% | Epoch AI | 2026-08-27 |
| Legal Research Bench | 23.1% | Vals AI | 2026-11-09 |
Frontend 52.4 ±0.7
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Vibe Code Bench v1.1 | 49.9% | Vals AI | 2026-11-09 |
| LMArena WebDev | 1,581 | LMArena | 2026-11-09 |
General 56.9 ±2.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| EQ-Bench 4 | 1,156 | EQ-Bench | 2026-10-24 |
| EQ-Bench Creative Writing v3 | 1,553 | EQ-Bench | 2026-11-11 |
| LMArena Text | 1,464 | LMArena | 2026-11-09 |
| Creative Story-Writing (Lech Mazur) | 57.0% | Lech Mazur | 2026-10-23 |
| EQ-Bench Longform Creative Writing | 75.6% | EQ-Bench | 2026-11-11 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| Chess Puzzles | 16.0% | Epoch AI | A stronger configuration result was used |
| Chess Puzzles | 4.0% | Epoch AI | A stronger configuration result was used |
| CritPt | 10.0% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond | 89.7% | Epoch AI | A stronger configuration result was used |
| GPQA Diamond | 73.2% | Epoch AI | A stronger configuration result was used |
| LMArena Text | 1,458 | LMArena | A stronger configuration result was used |
| LMArena WebDev | 1,448 | LMArena | A stronger configuration result was used |
| OTIS Mock AIME 2024-2025 | 95.6% | Epoch AI | A stronger configuration result was used |
| OTIS Mock AIME 2024-2025 | 46.7% | Epoch AI | A stronger configuration result was used |
| SciCode | 46.4% | Epoch AI | A stronger configuration result was used |
Names matched to this model
deepseek-v4-pro