Chronicles-OCR - Archaic Character Recognition: leaderboard

Chronicles-OCR (2026) tests vision-language models on historical Chinese writing across the Seven Chinese Scripts, 2,800 expert-annotated images from Oracle Bone to Cursive. This board scores recognition of a single archaic character highlighted by a box on Oracle Bone, Bronze and Seal images, as exact-match accuracy in percent averaged over the three scripts; higher is better.

Metric: Exact-match accuracy (%) of the modern character given for an archaic symbol highlighted by a box on the image, on the archaic scripts (Oracle Bone, Bronze and Seal), the paper Average column over the per-script results; higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 26 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Kimi K2.5Non-reasoning27.1#266
2Seed 2.0 ProNon-reasoning24.5#222
3Qwen 3.5 397B A17BNon-reasoning22.6#264
4Seed 1.8Non-reasoning20.6#252
5Gemini 3.1 Pro (Preview)19.5#99
6Qwen 3 VL 235B A22B Instruct17.5#532
7Qwen 3 VL 8B Instruct17.3#1029
8Qwen 3.5 35B A3BNon-reasoning16.2#492
9InternVL3.5-241B-A28B15.7#733
10Qwen 3 VL 235B A22B (Thinking)13.6#459
11Qwen 3 VL 2B Instruct10.7#2367
12Claude Opus 4.7Thinking10#85
13Qwen 3 VL 8B (Thinking)9.1#910
14Ovis2.6-30B-A3BThinking9
15MiMo-V2-Omni8.6#465

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

Interactive version: theaggregate.ai/benchmark?slug=chronicles-ocr-archaic-character-recognition · How It Works · Data refreshed daily, snapshot 2026-11-21.