Reframe Study - MME-RealWorld: leaderboard
Reframe v1 Table2, MME-RealWorld. Greedy native visual protocol: image5120 tokens/tool image1960; video64frames×128, selectedframe400/cumulative4096. Answer budgets8192 image/16384 video, max2tools/3turns. Qwen3.5-9B is no-tool; releasedEliYuan00/Reframe is separately trained. Strict match or Gemini3.1FlashLite adjudication; unparseable wrong. VideoMMEv2 Non-Lin, MLVU7-task mean, MVBench excludes fine-grainedpose;VSInative mixedaccuracy,V-STaRmAM×100. No derived tool-gain or unbound trained checkpoints.
Metric: Score (%). Source: arxiv.org. Saturation forecast: Around February 2027. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 9B | 71.45 | #803 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=reframe-study-mme-realworld · How It Works · Data refreshed daily, snapshot 2026-11-12.