M-GATE - Round-Trip Translation (Hard Subset): leaderboard
Round-trip English-to-target-to-English translation of M-GATE's fixed 50-sentence hard subset across 29 target languages, scored for meaning preservation by a three-provider small-model judge panel (arXiv 2608.03803v1, Appendix Table J.4). Mean normalized judge score from 0 to 1; higher is better. The evaluated model performs both translation directions.
Metric: Mean normalized judge score over 29 target languages (0-1; hard subset). Source: arxiv.org. Saturation forecast: Around February 2027. 82 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 (High) | 0.82 | #83 (GPT-5.4) |
| 2 | GPT-5.5 (High) | 0.82 | #33 (GPT-5.5) |
| 3 | Claude Fable 5 | 0.81 | #20 |
| 4 | GPT-5.5 (Medium) | 0.81 | #33 (GPT-5.5) |
| 5 | GPT-5.4 (Medium) | 0.8 | #83 (GPT-5.4) |
| 6 | GPT-5 | 0.78 | #105 |
| 7 | Claude Opus 4.6 (Thinking) | 0.78 | #67 (Claude Opus 4.6) |
| 8 | Gemini 3.1 Pro (Preview) | 0.77 | #64 |
| 9 | Gemini 3.5 Flash | 0.77 | #66 |
| 10 | Claude Opus 4.5 (Thinking) | 0.77 | #87 (Claude Opus 4.5) |
| 11 | Gemini 3.5 Flash (High) | 0.77 | #66 (Gemini 3.5 Flash) |
| 12 | GPT-5.4 Mini (High) | 0.77 | #228 (GPT-5.4 Mini) |
| 13 | GPT-5.2 (High) | 0.76 | #115 (GPT-5.2) |
| 14 | GPT-5.4 Mini (Medium) | 0.76 | #228 (GPT-5.4 Mini) |
| 15 | Gemini 2.5 Pro | 0.75 | #159 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=m-gate-round-trip-translation-hard-subset · How It Works · Data refreshed daily, snapshot 2026-11-05.