SPUR - Heterogeneous Integration: leaderboard

SPUR (2026) tests multimodal models with multiple-choice questions on multi-panel biomedical experimental figures from PubMed Central. Accuracy in percent on heterogeneous integration questions, which ask the model to combine evidence across panels of different types; higher is better.

Metric: Accuracy (%) on SPUR the 130 heterogeneous integration multiple-choice questions about multi-panel biomedical experimental figures from PubMed Central papers (questions drafted by GPT-4o, kept only when GPT-4o failed them in at least six of ten text-only attempts, then expert-reviewed): the questions ask the model to align and reason across panels of different types; higher is better. Source: arxiv.org. Saturation forecast: Not forecast. 19 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Ministral-3-14B-Instruct-251270#1639
2GLM-4.5V68.46#754
3Ministral-3-8B-Instruct-251266.15#1907
4Qwen 3 VL 30B A3B Instruct63.08#895
5Qwen 2.5 VL 72B Instruct61.9#841
6Gemini 2.5 Pro (Preview 06-05)61.54#213
7InternVL3-78B61.24#759
8Qwen 3 VL 30B A3B (Thinking)61.24#719
9Claude 3.7 SonnetThinking60.8#470
10Gemini 3 Pro (Preview)59.23#119
11O4 MiniHigh59.23#368
12Llama 4 Maverick58.46#1110
13Gemma 3 27B (IT)57.69#1324
14Mistral Small 3.156.92#1706
15Seed-1.656.92#469

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

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