A Claude Opus 5 debate asks whether bio classifiers are still necessary
TheZvi · x · 2026-07-25
Zvi argues Claude Opus 5 may make some safety filters look less necessary
The post reacts to a claim that Claude Opus 5 appears at least as good as Fable on virology-related tasks. The implication is that if Opus 5 can reach that level without Fable’s bio classifiers, it raises a question about why those extra classifiers are needed at all.
The reply also repeats a proposed explanation: the model may be doing too much self-verification and has poor task-scope calibration, but the author suggests those issues are likely fixable, especially if the model is faster and cheaper.
Related event: Claude Opus 5 Virology Capabilities Spark Safety Debate(2 posts)→
More from Models
- Release blog teaser shows a near-tie on FrontierCode agentic coding benchmark — hardmaru · 2026-07-25
- User Accuses Anthropic of Gaming ARC-AGI-3 by Training Specifically on Benchmark Patterns — VraserX · 2026-07-25
- Moonshot’s open-weight Kimi K3 is nearing top U.S. models and shifting AI economics — OmarUFlorez · 2026-07-25
- Opus 5 tops nine biology benchmarks, but still trails in some analysis tasks — kenbwork · 2026-07-25
- Claude Opus 5 now allows source-code vulnerability discovery but still blocks binaries — TheZvi · 2026-07-25
- Thread argues scaled RL and large distillation favor Anthropic in the current race — rickasaurus · 2026-07-25