DeepSeek V4 Flash 0731
| Released | 2026-07-31 |
|---|---|
| Price per 1M | $0.15 in · $0.30 out |
| Overall score | 56.5 · rank 62 |
| Context window | 1,000,000 tokens |
| Provider | DeepSeek |
| Weights | Open weights (MIT) |
Scores
| Dimension | Score | Rank | Benchmarks |
|---|---|---|---|
| Overall: Balanced | 56.5 | 62 | 30 |
| Overall: Coding | 56.8 | 59 | 21 |
| Overall: Reasoning | 57.8 | 59 | 15 |
| Overall: Agentic work | 58.2 | 52 | 20 |
| Overall: Frontend | 56.7 | 50 | 7 |
| Coding | 55.2 | 70 | 6 |
| Agents | 59.1 | 39 | 6 |
| Reasoning | 60.1 | 44 | 8 |
| Math | 57.5 | 47 | 4 |
| Knowledge | 53.1 | 77 | 3 |
| Frontend | 57.5 | 41 | 1 |
| General | 50.9 | 93 | 2 |
| Value vs price | +13.7 | 13 |
Configurations
| Configuration | Balanced | Benchmarks |
|---|---|---|
| Unspecified setting | 56.4 ±3.6 | 5 |
| High | 57.4 ±1.5 | 18 |
| Low | 58.7 ±5.7 | 2 |
| Max | 56.1 ±3.3 | 11 |
| Non-reasoning | 45.7 ±5.7 | 2 |
Evidence
Coding 55.2 ±2.4
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierCode | 18.8% | Epoch AI | 2026-11-11 |
| IOI | 32.7% | Vals AI | 2026-10-29 |
| SWE-bench Verified (Vals run) | 88.8% | Vals AI | 2026-09-01 |
| SciCode | 49.9% | Epoch AI | 2026-11-11 |
| LiveCodeBench (Vals run) | 87.3% | Vals AI | 2026-09-01 |
| ProgramBench | 0.0% | Vals AI | 2026-11-09 |
Agents 59.1 ±1.3
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Terminal-Bench 4.0 (Vals run) | 18.7% | Vals AI | 2026-11-09 |
| Terminal-Bench-Science | 4.3% | Vals AI | 2026-11-09 |
| Terminal-Bench 2.1 (Vals run) | 67.0% | Vals AI | 2026-10-01 |
| Tax Agent Bench | 28.1% | Vals AI | 2026-11-09 |
| SkillsBench | 50.7% | Vals AI | 2026-10-29 |
| CyberBench v1.1 | 72.1% | Vals AI | 2026-11-05 |
Reasoning 60.1 ±1.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| ARC-AGI-2 | 61.4% | Epoch AI | 2026-11-11 |
| ARC-AGI-2 (ARC Prize verified) | 61.4% | ARC Prize | 2026-11-11 |
| CritPt | 16.6% | Epoch AI | 2026-11-11 |
| ARC-AGI-1 (ARC Prize verified) | 89.0% | ARC Prize | 2026-11-11 |
| ARC-AGI-1 | 89.0% | Epoch AI | 2026-11-11 |
| SimpleBench | 61.1% | Epoch AI | 2026-11-11 |
| GPQA Diamond | 91.0% | Epoch AI | 2026-08-02 |
| GPQA Diamond (Vals run) | 89.9% | Vals AI | 2026-09-01 |
| BioMysteryBench | 64.4% | Vals AI | 2026-11-09 |
| Chess Puzzles | 32.0% | Epoch AI | 2026-10-13 |
| Mystery Game Puzzles | 34.0% | Epoch AI | 2026-08-05 |
Math 57.5 ±1.2
| Benchmark | Result | Source | Date |
|---|---|---|---|
| FrontierMath Tiers 1-3 | 57.5% | Epoch AI | 2026-08-02 |
| FrontierMath Tier 4 | 24.4% | Epoch AI | 2026-08-02 |
| OTIS Mock AIME 2024-2025 | 94.4% | Epoch AI | 2026-08-02 |
| ProofBench v1.1 | 56.0% | Vals AI | 2026-11-09 |
| ProofBench | 56.0% | Epoch AI | 2026-11-11 |
Knowledge 53.1 ±3.6
| Benchmark | Result | Source | Date |
|---|---|---|---|
| SimpleQA Verified | 33.6% | Epoch AI | 2026-08-27 |
| MMLU-Pro (Vals run) | 86.2% | Vals AI | 2026-09-01 |
| Legal Research Bench | 30.3% | Vals AI | 2026-11-09 |
Frontend 57.5 ±0.8
| Benchmark | Result | Source | Date |
|---|---|---|---|
| Vibe Code Bench v1.1 | 74.7% | Vals AI | 2026-11-09 |
General 50.9 ±3.0
| Benchmark | Result | Source | Date |
|---|---|---|---|
| EQ-Bench Creative Writing v3 | 1,441 | EQ-Bench | 2026-11-11 |
| EQ-Bench Longform Creative Writing | 61.2% | EQ-Bench | 2026-11-11 |
Results not used in the score
| Benchmark | Result | Source | Why not used |
|---|---|---|---|
| ARC-AGI-1 | 87.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 | 84.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 | 11.8% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 87.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 84.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-1 (ARC Prize verified) | 11.8% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 | 56.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 | 46.0% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 | 2.1% | Epoch AI | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 56.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 46.0% | ARC Prize | A stronger configuration result was used |
| ARC-AGI-2 (ARC Prize verified) | 2.1% | ARC Prize | A stronger configuration result was used |
Names matched to this model
deepseek-v4-flash-0731