PBSBench - Cell True/False: leaderboard

PBSBench (2026) tests vision-language models on peripheral blood smear hematopathology, with questions about single white blood cell crops. True or false accuracy in percent on in-domain cells; higher is better.

Metric: Accuracy (%, times 100) on PBSBench cell-level in-domain true/false questions about single white-blood-cell crops from peripheral blood smears (957 questions); questions generated from GPT-4o morphology captions of hematopathologist-labeled cells; greedy decoding, model chat templates; higher is better. Source: arxiv.org. Saturation forecast: Around June 2028. 16 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-4o77#754
2InternVL3.5-8B73#1618
3Gemini 2.5 Pro70#280
4Qwen 2 VL 7B Instruct67#2581
5PathGen-LLaVA65#1421
6Qwen 2.5 VL 7B Instruct63#1919
7DeepSeek-VL-7B63#7442
8Quilt-LLaVA61#5009
9Claude Sonnet 4.559#267
10Idefics3-8B-Llama359#5592
11llava-v1.6-mistral-7B59#4027
12LLaVA-1.6-Vicuna-7B58
13LLaVA-Med54
14BLIP2-OPT-6.7B51
15Med-Flamingo45

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=pbsbench-cell-true-false · How It Works · Data refreshed daily, snapshot 2026-11-22.