Gemini 4 Argon Hits 1.73% WER on OCR Benchmark, Surpassing Human Experts
gabriberton · x · 2026-10-28
OCR is the latest task where AI surpassed humans — and it fell before math. Gemini 4 Argon achieves clear SOTA on a transcription benchmark: 1.73% modified WER and 0.64% CER at low effort, with the benchmark far from saturated.
- Argon follows instructions better, distinguishing and ignoring archival markings and page numbers, and handles unusually formatted pages well
- Its errors are far less consequential than human experts', to the point the author argues they shouldn't even be called errors: "on our benchmark anyway, Argon is much better than human experts."
More from Models
- Debate: Claude isn't 'lobotomized' — inference stack updates may shift old-model outputs — JasonBotterill · 2026-10-28
- AI labs near IPO, yet LLMs still produce 'astonishingly terrible prose' by default — GabGarrett · 2026-10-28
- Rebuttal: Kimi 3.1 Nears Fable on Frontier Code 2, Contradicting 'China Models Crushed' Claim — bookwormengr · 2026-10-28
- Open model Flint hits #2 trending on Hugging Face, team outlines next steps — josh_tobin_ · 2026-10-28
- Artificial Analysis Intelligence Index v5 widens open-weight gap from 3 to 12 points — ResearchCrafty1804 · 2026-10-28
- YacineMTB rails: even the best LLMs are 'complete dogshit' and he's tired of it — yacineMTB · 2026-10-28