Opus 5.2 spotted in Claude Code routing; leaked Model 2 reportedly writes most of Anthropic's code
新智元 · wechat · 2026-09-15
- Developers found Claude Code's "Opus5" silently routes to Opus 5.2 — apparently skipping 5.1. Reports: faster responses, cleaner output, and self-driven "gauntlet loops" that iterate code without laziness.
- A community "Tibo probe" (asking about an obscure figure with search off) distinguishes routed 5.2 sessions from old weights.
- A leaked August internal risk report claims an unreleased "Model 2" scored 62.8% on CoBench v2 (12.5 points above Mythos5) and already writes most of Anthropic's company code via agents; an RSI model reportedly scores 22 points higher still, said to replace 85% of research-team work.
Related event: Anthropic appears to be quietly testing Opus 5.2 via Claude Code(6 posts)→
More from Models
- Leak claims GPT-6.1 sol saturated FrontierMath Tier 4 and solved 5 Erdős problems in 10 months — haider1 · 2026-10-03
- Anthropic's 'reasoning extraction' blocks on Claude spark user backlash — cephaloform · 2026-10-03
- "GPT-7 Won't Be Made by Humans": Viral Account Fuels AI-Self-Improvement Speculation — iruletheworldmo · 2026-10-03
- Fan praises Anthropic for shipping solid cheap models without hype pre-marketing — rsuyoy · 2026-10-03
- Custom 200-Task Benchmark: Gemma-4-26B Tops Tier List Even Quantized to iq3s — Potential-Net-9375 · 2026-10-03
- Heavy user: Opus 5.5 closed the instruction-following gap and now handles most of my coding — brandon_galang · 2026-10-03