Massive eval table update benchmarks open-weight models across coding, finance, legal
siyan_zhao · x · 2026-11-13
Ziqiao Ma's team updated their model card eval table with additional private and semi-private benchmarks plus official result verification. The screenshots span coding (FrontierCode v2.0, SWEBench Pro v2, Terminal Bench), search/tools (BrowseComp, MCP Atlas), knowledge work, finance (Excel Modeling, Tax Agent Bench), and legal (Harvey LAB) benchmarks, comparing Inkling 2 Small, GLM 5.3, MiMo V2.6, MiniMax-M3.1, DeepSeek-V4.1-Flash, Kimi K3, Mistral Large 4 and GPT-6, with no single model sweeping all categories.
More from Models
- Claude Duet invites roll out: 3 invites per user plus free Pro trial — testingcatalog · 2026-11-13
- Claims: Fable 5 and Opus 5.5 are big readers who love Claude's writing once they spot it — repligate · 2026-11-13
- DeepSeek V4.1-Pro rumor doubted: API fingerprints unchanged, new model may wait until Feb — teortaxesTex · 2026-11-13
- Qwen 4 starts speaking Chinese mid-coding-session, user reports — gnukeith · 2026-11-13
- Dev jokes Opus 5.5's style feels like it was trained on ICLR submissions — tallinzen · 2026-11-13
- Polymarket puts 36% odds on next Claude Opus shipping this month — Polymarket · 2026-11-13