Sarvam Vision 2.1 tops olmOCR-Bench at 87.3, beating Opus 5's 85.1
HarveenChadha · x · 2026-09-24
Sarvam AI claims its vision model Sarvam Vision 2.1 scores 87.3 on olmOCR-Bench, ahead of every model evaluated including Infinity-Parser2 Pro (86.1) and Opus 5 (85.1). The benchmark measures pass-fail accuracy across arXiv math, old scans, tables, multi-column pages and dense text — a representative test of document parsing. If it holds up, a smaller team's model beating top general models on specialized OCR is notable (self-reported, pending independent verification).
Related event: Sarvam Vision 2.1 Tops OCR Benchmarks Across 22 Indian Languages(3 posts)→
More from Models
- LiquidAI extends lossless speculative decoding to vision-language models — JosephJacks_ · 2026-09-25
- GPT-5.2 solves a COLT 2022 open problem the researcher had chased since 2016 — kfountou · 2026-09-25
- Agents bypass monitoring guardrails with strategies that improve as reasoning effort scales — maksym_andr · 2026-09-25
- micro1 launches flow-transform 1.0, hits 96.0% F1 on PrivacyBench PII transformation — omarsar0 · 2026-09-25
- UkisAI ships Swift reasoning LLM family: -63.4% thinking tokens at 1.8x speed on Qwen base — Secure_Recording_472 · 2026-09-25
- "How Many Versions of Humanity's Last Exam Before They Rename It?" — code_star · 2026-09-25