AI lab · founded 2023 · 14 models tracked
DeepSeek models
Every DeepSeek model on the API today, with release dates, context windows, API prices and benchmark scores. Newest: DeepSeek V4.1 Flash 1017, 5d ago.
- First model
- Nov 2, 2023 DeepSeek Coder
- Models · 30 days
- 1 of 14
- Latest release
- 5d ago
- Cheapest input / 1M
- $0.047 DeepSeek V4 Flash 0731
- Largest context
- 1M DeepSeek V4.1 Flash 1017
Release history
Timeline
DeepSeek ships a new model about every 51 days. Each dot is a release; click one to open it. Hollow dots are retired models.
Lineup
All DeepSeek models
Benchmarks
DeepSeek's best scores
The top DeepSeek model on each benchmark, and where it ranks among all 281 tracked models.
| Benchmark | Best DeepSeek model | Score | Overall rank |
|---|---|---|---|
| FrontierCodeSoftware engineering | DeepSeek V4.1 Flash 1017 | 43.9% | #16of 41 |
| SWE-bench VerifiedSoftware engineering | DeepSeek V4 Pro 0813 | 77.6% | #5of 32 |
| Terminal-BenchSoftware engineering | DeepSeek V3.2 | 39.6% | #23of 40 |
| SciCodeSoftware engineering | DeepSeek V4.1 Flash 1017 | 51.8% | #39of 102 |
| Aider PolyglotSoftware engineering | DeepSeek V3.2 | 74.2% | #4of 26 |
| WeirdML v3Software engineering | DeepSeek V4.1 Flash 1017 | 5.9% | #18of 22 |
| METR time horizonSoftware engineering | R1 0528 | 31 min | #16of 23 |
| APEX-AgentsAgents & computer use | DeepSeek V4 Pro 0813 | 47.3% | #30of 51 |
| Vending-Bench 2Agents & computer use | DeepSeek V4 Pro 0813 | $3,285 | #41of 61 |
| Lech Mazur WritingWriting & content | R1 0528 | 8.30/10 | #8of 27 |
| Fiction.LiveBenchWriting & content | DeepSeek V3.2 | 83.3% | #5of 24 |
| DeepResearch BenchWriting & content | R1 0528 | 35.1% | #20of 21 |
| GDP.pdfWriting & content | DeepSeek V4.1 Flash 1017 | 19.8% | #22of 38 |
| SimpleQA VerifiedWriting & content | DeepSeek V4 Pro 0813 | 52.9% | #22of 79 |
| WebDev ArenaDesign & visual | DeepSeek V4.1 Flash 1017 | 1614 | #18of 90 |
| Furniture AssemblyDesign & visual | DeepSeek V4.1 Flash 1017 | 34.2% | #20of 34 |
| GPQA DiamondReasoning & knowledge | DeepSeek V4.1 Flash 1017 | 93.2% | #17of 122 |
| ARC-AGI-2Reasoning & knowledge | DeepSeek V4.1 Flash 1017 | 72.9% | #19of 75 |
| ARC-AGI-1Reasoning & knowledge | DeepSeek V4.1 Flash 1017 | 94.5% | #16of 75 |
| SimpleBenchReasoning & knowledge | DeepSeek V4 Flash 0731 | 61.1% | #18of 58 |
| CritPtReasoning & knowledge | DeepSeek V4 Pro 0813 | 18.0% | #29of 111 |
| Chess PuzzlesReasoning & knowledge | DeepSeek V4.1 Flash 1017 | 48.0% | #17of 83 |
| LMCAReasoning & knowledge | DeepSeek V4.1 Flash 1017 | 47.0% | #32of 123 |
| FrontierMathMaths | DeepSeek V4.1 Flash 1017 | 71.2% | #25of 87 |
| FrontierMath Tier 4Maths | DeepSeek V4.1 Flash 1017 | 46.3% | #19of 64 |
| ProofBenchMaths | DeepSeek V4 Flash 0731 | 56.0% | #24of 70 |
| OTIS Mock AIMEMaths | DeepSeek V4.1 Flash 1017 | 99.4% | #12of 120 |
Directory