University of Basel RISE · Research and Infrastructure SupportHumanities Data Benchmark

stepfun/step-3.7-flash

StepFunvia OpenRouter
average±2.1 (95%)
72.7
leaderboard rankof 114
≈54
benchmarksof 12
12
per 1,000 documentsmean over its benchmarks
$10
seconds per inputmean response time
99.7
Compare with other models Its row in the Matrix All its test runs

By task

  • Data correction

    68.0

    Best of 114 models with all its benchmarks: claude-opus-5-5, 82.7.

  • Document understanding

    61.8

    Best of 107 models with all its benchmarks: gpt-6-astra, 97.0.

  • Information extraction

    81.3

    Best of 107 models with all its benchmarks: grok-4.6, 90.8.

  • Transcription

    83.8

    Best of 114 models with all its benchmarks: gemini-4-argon, 96.2.

By benchmark

BenchmarkScoreThe 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 docss / inputBest run
Bibliographic Data 96.4±1.7 ≈40of 115 $15$0.076 run 110.96 T1128 · 5 Jun 2026prompt.txt
Blacklist Cards 84.0±4.4 ≈89of 114 $4.80$0.16 run 25.43 T1135 · 17 Sep 2026prompt.txt
Book Advert XML Repair 40.4±10.1 ≈16of 119 $20$0.99 run 103.66 T1138 · 17 Sep 2026prompt.txt
Business Letters 60.0±6.1 ≈24of 114 $3.69$0.21 run 24.42 T1129 · 17 Sep 2026prompt.txt
Company Lists 89.3±6.2 ≈6of 114 $6.77$0.10 run 30.89 T1136 · 17 Sep 2026prompt.txt
Duty Rosters 77.8±4.5 ≈46of 107 $17$0.36 run 116.05 T1340 · 22 Jul 2026prompt.txt
Fraktur Adverts 93.1±1.1 ≈32of 114 $19$0.094 run 476.11 T1132 · 5 Jun 2026prompt_optimized.txt
General Meeting Minutes 78.3±13.2 ≈25of 108 $7.90$0.071 run 95.25 T1271 · 3 Jul 2026prompt.txt
Library Cards 83.3±1.7 ≈81of 114 $3.44$0.90 run 19.53 T1133 · 17 Sep 2026prompt.txt
Magazine Pages 12.0±9.2 ≈66of 115 $5.53$0.25 run 80.75 T1140 · 5 Jun 2026prompt.txt
Medieval Manuscripts 62.7±12.9 ≈72of 114 $11$0.14 run 81.09 T1134 · 17 Sep 2026prompt.txt
Personnel Cards 95.5±1.4 ≈58of 114 $6.63$0.40 run 32.42 T1139 · 17 Sep 2026prompt.txt

Price

FromInput, $ per million tokensOutput, $ per million tokensSource
5 Jun 20260.21.15archived price page

Every run

  1. T1340Duty Rosters17 Sep 202674.1$0.37
  2. T1271General Meeting Minutes17 Sep 202673.6$0.066
  3. T1140Magazine Pages17 Sep 20263.3$0.25
  4. T1139Personnel Cards17 Sep 202695.5$0.40
  5. T1138Book Advert XML Repair17 Sep 202640.4$0.99
  6. T1137Company Lists17 Sep 202656.2$0.079
  7. T1136Company Lists17 Sep 202689.3$0.10
  8. T1135Blacklist Cards17 Sep 202684.0$0.16
  9. T1134Medieval Manuscripts17 Sep 202662.7$0.14
  10. T1133Library Cards17 Sep 202683.3$0.90
  11. T1132Fraktur Adverts17 Sep 202692.0$0.084
  12. T1131Business Letters17 Sep 2026 · 32 signed letters61.0$0.14
  13. T1130Business Letters17 Sep 2026 · 25 unsigned letters57.0$0.084
  14. T1129Business Letters17 Sep 2026 · all 57 letters60.0$0.21
  15. T1128Bibliographic Data17 Sep 202686.7$0.075
  16. T1340Duty Rosters22 Jul 202677.8$0.36
  17. T1271General Meeting Minutes3 Jul 202678.3$0.071
  18. T1140Magazine Pages5 Jun 202612.0$0.25
  19. T1139Personnel Cards5 Jun 202695.2$0.42
  20. T1138Book Advert XML Repair5 Jun 202633.3$1.25
  21. T1137Company Lists5 Jun 202658.8$0.072
  22. T1136Company Lists5 Jun 202679.3$0.13
  23. T1135Blacklist Cards5 Jun 202683.9$0.17
  24. T1134Medieval Manuscripts5 Jun 202660.8$0.14
  25. T1133Library Cards5 Jun 202683.2$0.85
  26. T1132Fraktur Adverts5 Jun 202693.1$0.094
  27. T1131Business Letters5 Jun 2026 · 32 signed letters62.0$0.14
  28. T1130Business Letters5 Jun 2026 · 25 unsigned letters59.0$0.083
  29. T1129Business Letters5 Jun 2026 · all 57 letters59.0$0.22
  30. T1128Bibliographic Data5 Jun 202696.4$0.076