HM-Bench (PCA): leaderboard
HM-Bench (2026) tests multimodal LLMs on hyperspectral image understanding with 19,337 multiple-choice questions over 2,178 samples in 13 perception and reasoning tasks, given PCA-only input: a false-colour image of the cube's leading principal components. Accuracy (%), zero-shot; higher is better.
Metric: Average accuracy (%) over all tasks of HM-Bench's 19,337 hyperspectral multiple-choice questions (2,178 samples, 13 task categories) with PCA-only input: a false-colour image of the cube's leading principal components; zero-shot, one option letter as output, temperature 0, at most 64 new tokens; higher is better. Source: arxiv.org. Saturation forecast: Around 2032. 16 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | InternVL3.5-14B | 43.08 | #1504 |
| 2 | GPT-5.4 Mini | 42.35 | #406 |
| 3 | InternVL3-9B | 42.14 | #1728 |
| 4 | GLM-4.6V-Flash | 42.06 | #1040 |
| 5 | Qwen 3 VL 4B | 40.96 | #1151 |
| 6 | Claude Sonnet 4.6 | 39.48 | #173 |
| 7 | LLaVA-Next-8B | 39.03 | #2046 |
| 8 | Kimi-VL-A3B-Instruct | 38.93 | #2253 |
| 9 | Llama-3.1-Nemotron-Nano-VL-8B-V1 | 38.67 | |
| 10 | InternVL2-8B | 38.09 | #3227 |
| 11 | InternVL3-14B | 38.09 | #1214 |
| 12 | GeoLLaVA-8K | 37.79 | |
| 13 | GeoChat | 35.27 | #3155 |
| 14 | InternVL2.5-4B | 34.83 | #2549 |
| 15 | Grok 4 | 32.78 | #319 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=hm-bench-pca · How It Works · Data refreshed daily, snapshot 2026-11-21.