HalluHard: Sonnet 5 and Fable 5 Show No Major Drop in Hallucinations
maksym_andr · x · 2026-07-06
Latest data from the HalluHard hallucination benchmark shows Sonnet 5 at a 50.9% hallucination rate and Fable 5 at 59.7%, with both performing better without web search than with it. Compared to GLM-5.2's 74.8%, the Claude series leads overall, though researchers note no significant breakthrough.
The author criticizes the AI community for hyper-focusing on Agent capabilities while neglecting hallucination issues that impact almost all real users, urging the industry to reprioritize.
More from Models
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- Meta's Muse Agent has built-in invite code logic, hinting at free-usage expansion — testingcatalog · 2026-09-11
- Same Echo Maze prompt, three frontier models: all passed visually but shipped the same hidden bug — eyishazyer · 2026-09-11
- Benchmark scores drop from 89% to 19% on new evals — how benchmaxxing breaks leaderboard trust — airesearch12 · 2026-09-11
- ChatGPT tells user their question is too hard and to 'accept dumber answers' — phido3000 · 2026-09-11
- Claude is no longer available for minors as Anthropic rolls out age assurance — Muhammad523 · 2026-09-11