Cross-model test says latest Claude refuses a prompt that many frontier models answer
cyb3rops · x · 2026-07-23
A screenshot-based comparison claims that a Hugging Face/OpenAI incident is being over-framed as FUD, and that the latest Claude models refuse to help on a prompt that many other frontier models will answer.
- The author says they ran the same prompt across every model available in OpenRouter and published the results.
- The linked example suggests the latest Claude models refuse, while many other models — including recent OpenAI ones — provide assistance.
- The screenshot shows a benchmark-style dashboard with 342 models, where one model (moonshotai/kimi-latest) is shown as a 10/10 result in the displayed case.
The post is essentially about cross-model safety/refusal behavior on a specific task, and the debate over how much of the incident is model safety versus presentation bias.
More from Models
- Kimi K3.1 Pro tops MathArena open models, closing frontier gap to about 6 months — teortaxesTex · 2026-10-26
- Ruff creator Charlie Marsh: frontier models find infinite bugs in any software — charliermarsh · 2026-10-26
- Dev testing one skill across Opus 5.5 and Sol 6.1 finds instruction following 'all over the map' — lucasmeijer · 2026-10-26
- supermemory unveils learner-1, claiming continual learning for any model — julianweisser · 2026-10-26
- Anthropic says Opus 5.5 cuts cost per task by ~40% vs Opus 5 — taufiqintech · 2026-10-26
- AI search APIs ranked: top 12 within 7 points, yet prices differ nearly 7x — FinanceYF5 · 2026-10-26