Skip to content
TifinaghBench

Compare runs

Pick two to four runs. The link updates as you choose, so you can share the comparison.

Runs to compare

GPT-6.1-SolhighGPT-6-AstralowClaude Opus 5.5high

Head to head

Differences in TifinaghBench score, with the 95% interval of the difference.

ModelModelDifferenceVerdict
GPT-6.1-Sol 29.9GPT-6-Astra 29.4+0.5 ± 1.5Statistical tie
GPT-6.1-Sol 29.9Claude Opus 5.5 27.3+2.6 ± 1.4GPT-6.1-Sol ahead
GPT-6-Astra 29.4Claude Opus 5.5 27.3+2.1 ± 1.4GPT-6-Astra ahead

Every cell, side by side

Scores out of 100 in each dialect and task; bars share one scale.

  • GPT-6.1-Sol
  • GPT-6-Astra
  • Claude Opus 5.5

Meaning

Standard Moroccan Amazigh · Tifinagh → English
  • GPT-6.1-Sol: 31.0
  • GPT-6-Astra: 29.0
  • Claude Opus 5.5: 22.0
Standard Moroccan Amazigh · English → Tifinagh
  • GPT-6.1-Sol: 11.2
  • GPT-6-Astra: 11.7
  • Claude Opus 5.5: 9.6
Tarifit · Tifinagh → English
  • GPT-6.1-Sol: 38.7
  • GPT-6-Astra: 37.8
  • Claude Opus 5.5: 37.2
Tarifit · English → Tifinagh
  • GPT-6.1-Sol: 8.0
  • GPT-6-Astra: 9.0
  • Claude Opus 5.5: 7.0
Kabyle · Tifinagh → English
  • GPT-6.1-Sol: 62.0
  • GPT-6-Astra: 62.1
  • Claude Opus 5.5: 59.0
Kabyle · English → Tifinagh
  • GPT-6.1-Sol: 28.4
  • GPT-6-Astra: 26.7
  • Claude Opus 5.5: 29.0

Script

Standard Moroccan Amazigh · Tifinagh ↔ Latin
  • GPT-6.1-Sol: 99.6
  • GPT-6-Astra: 99.0
  • Claude Opus 5.5: 99.1
Tarifit · Tifinagh ↔ Latin
  • GPT-6.1-Sol: 92.8
  • GPT-6-Astra: 93.0
  • Claude Opus 5.5: 92.5
Kabyle · Tifinagh ↔ Latin
  • GPT-6.1-Sol: 97.0
  • GPT-6-Astra: 97.1
  • Claude Opus 5.5: 96.5