deepseek-flash
DeepSeek- average±2.0 (95%)
- 82.9
- leaderboard rankof 114
- ≈14
- benchmarksof 12
- 12
- per 1,000 documentsmean over its benchmarks
- $13
- seconds per inputmean response time
- 45.1
By task
-
Data correction
73.8
Best of 114 models with all its benchmarks: claude-opus-5-5, 82.7.
-
Document understanding
84.4
Best of 107 models with all its benchmarks: gpt-6-astra, 97.0.
-
Information extraction
86.3
Best of 107 models with all its benchmarks: grok-4.6, 90.8.
-
Transcription
86.5
Best of 114 models with all its benchmarks: gemini-4-argon, 96.2.
By benchmark
| Benchmark | ScoreThe best current run on each benchmark, on the benchmark's metric scaled to 0–100. ± is the 95% sampling margin, from resampling the run's inputs. | RankRank among all current models on that benchmark; ≈ for models that cannot be told apart within sampling error. | Per 1,000 docs | s / input | Best run |
|---|---|---|---|---|---|
| Bibliographic Data | 97.8±1.7 | ≈8of 115 | $14$0.069 run | 44.27 | T1765 · 16 Sep 2026prompt.txt |
| Blacklist Cards | 95.7±2.2 | ≈14of 114 | $4.77$0.16 run | 18.14 | T1772 · 16 Sep 2026prompt.txt |
| Book Advert XML Repair | 53.7±11.4 | ≈1of 119 | $15$0.74 run | 49.1 | T1775 · 16 Sep 2026prompt.txt |
| Business Letters | 62.0±7.2 | ≈24of 114 | $4.69$0.27 run | 17.94 | T1766 · 16 Sep 2026prompt.txt |
| Company Lists | 95.1±2.7 | ≈1of 114 | $8.15$0.12 run | 27.39 | T1773 · 16 Sep 2026prompt.txt |
| Duty Rosters | 81.6±3.2 | ≈30of 107 | $35$0.73 run | 119.11 | T1779 · 16 Sep 2026prompt.txt |
| Fraktur Adverts | 95.5±1.7 | ≈32of 114 | $19$0.093 run | 68.49 | T1769 · 16 Sep 2026prompt_optimized.txt |
| General Meeting Minutes | 85.5±9.0 | ≈25of 108 | $19$0.17 run | 65.51 | T1778 · 16 Sep 2026prompt.txt |
| Library Cards | 86.1±1.5 | ≈51of 114 | $2.27$0.60 run | 8.15 | T1770 · 16 Sep 2026prompt.txt |
| Magazine Pages | 77.6±10.8 | ≈18of 115 | $6.66$0.31 run | 27.03 | T1777 · 16 Sep 2026prompt.txt |
| Medieval Manuscripts | 70.0±12.2 | ≈27of 114 | $21$0.25 run | 76.71 | T1771 · 16 Sep 2026prompt.txt |
| Personnel Cards | 93.9±2.3 | ≈58of 114 | $6.33$0.39 run | 19.82 | T1776 · 16 Sep 2026prompt.txt |
Price
| From | Input, $ per million tokens | Output, $ per million tokens | Source |
|---|---|---|---|
| 16 Sep 2026 | 0.3 | 1.2 | archived price page |
Every run
- T1765Bibliographic Data26 Sep 202697.7$0.076
- T1779Duty Rosters16 Sep 202681.6$0.73
- T1778General Meeting Minutes16 Sep 202685.5$0.17
- T1777Magazine Pages16 Sep 202677.6$0.31
- T1776Personnel Cards16 Sep 202693.9$0.39
- T1775Book Advert XML Repair16 Sep 202653.7$0.74
- T1774Company Lists16 Sep 202663.3$0.11
- T1773Company Lists16 Sep 202695.1$0.12
- T1772Blacklist Cards16 Sep 202695.7$0.16
- T1771Medieval Manuscripts16 Sep 202670.0$0.25
- T1770Library Cards16 Sep 202686.1$0.60
- T1769Fraktur Adverts16 Sep 202695.5$0.093
- T1768Business Letters16 Sep 2026 · 32 signed letters66.0$0.18
- T1767Business Letters16 Sep 2026 · 25 unsigned letters58.0$0.12
- T1766Business Letters16 Sep 2026 · all 57 letters62.0$0.27
- T1765Bibliographic Data16 Sep 202697.8$0.069