Tax Agent Bench shows best LLM scores under 50% on tax questions
ziv_ravid · x · 2026-11-01
The author gave the same tax issue to several LLMs and got conflicting answers. Searching around, they found a dedicated Tax Agent Bench — and even the best-performing model (Fable 5.1) gets under 50% of questions right. A reminder that LLMs remain unreliable for professional tax work without human verification.
More from Models
- Qwen3.8-Flash-Next under fire: strong at coding, falls apart on follow-up questions in research use — rm-rf-rm · 2026-11-01
- Dev claims Claude Opus 5.5 does in 1 day what 28 days of Unlimited Sol can't — karmay007 · 2026-11-01
- Users report ChatGPT chats drawing on Codex credits, sparking confusion — rolyataylor2 · 2026-11-01
- Small models as first-pass filters: Chris Albon's weekly scan over hundreds of thousands of entities — chrisalbon · 2026-11-01
- davidad podcast on emergent wisdom in LLMs, says Anthropic should stop cyber-offense training — RichardMCNgo · 2026-11-01
- Leaked OpenAI email hints ChatGPT will merge Chat and Work, unify usage limits — Angaisb_ · 2026-11-01