Unverified50% confidenceReleaseExact time
Alibaba's ATH team released the 0.8B-parameter unified document model Ovis OC23, which reached 96.74 on OmniDocBench, ranked first on PureDocBench with 82.31, and took gold on both OCRBench leaderboards.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/29/2026
First Seen
Valid until: 1/27/2027
Sources
Related Entities
Related Claims
UnverifiedQwen3.8-27B scored 91.1 on OmniDocBench v1.5, the strongest document parsing score among the generative models in this evaluation.72% similarUnverifiedScores across 19 models ranged from 0.55 to 1.00, with Qwen 3 235B A22B Instruct ranking first at 1.00.61% similarUnverifiedIn the Codeforces rating, o1's score is far lower than that of o3-mini-high.61% similarUnverifiedThe top of SWE-bench_Verified is ornith-ai/Ornith-1.5-397B with a score of 86; SWE-bench_Pro is taken by inclusionAI/Ling-3.1-flash with 68.74, and this model also ranks first on SWE-bench_Multilingual (88.5) and skillsbench (68.7).60% similarUnverifiedUnlimited OCR 2 scored 77.1 on MarathonDocBench.60% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1111450API
curl https://kongchang.com/api/v1/knowledge/claims/1111450MCP
get_claim(id=1111450)