DDx-TRACE - Localization: leaderboard

DDx-TRACE (2026) scores diagnostic trajectories on physician-adjudicated neuroradiology cases: models request imaging in free form, update a probabilistic differential diagnosis after each turn and stop with a localized diagnosis, judged against case rubrics. This board scores the final lesion localization, on a 0 to 1 scale; higher is better.

Metric: Final lesion localization score (0-1: laterality, region and substructure) on DDx-TRACE (211 physician-adjudicated multimodal neuroradiology cases), the model starts from limited history, requests imaging in free form (at most 6 requests), updates a four-item probabilistic differential after each turn and stops with a localized diagnosis, provider default decoding; endpoints judged against case rubrics by Gemini 3 Flash at temperature 0; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 13 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.40.621#135
2Gemini 3 Flash0.586#175
3Gemini 3.1 Pro (Preview)0.578#99
4Claude Sonnet 4.60.519#173
5GPT-5.4 Mini0.482#406
6Qwen 3.5 35B A3B0.472#492
7Qwen 3.5 27B0.439#375
8Qwen 3.5 9B0.365#828
9Lingshu-32B0.3#1113
10Qwen 3.5 4B0.282#1241
11Gemma 3 27B0.272#1663
12MedGemma-27B-IT0.264#1350
13MedGemma 1.5 4B0.187#2716

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=ddx-trace-localization · How It Works · Data refreshed daily, snapshot 2026-11-21.