Anthropic unveils Opus 4.8, boosting honesty and coding
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Anthropic released Opus 4.8, its latest Claude model, at the same price as Opus 4.7.
- Opus 4.8 adds a dynamic workflow tool that improves self‑knowledge and flags uncertainties.
- The model is described as more honest, owning what it doesn’t know and flagging problems in its own code.
- Opus 4.8 delivers stronger coding performance, beating GPT‑5.5 on a benchmark with 69.2% SWE.
- Anthropic is preparing a broader Mythos‑level release in the coming weeks, with Opus 4.8 as a preview.
- Developers can activate a high‑effort mode via /effort for more thorough results, and rate limits were raised to cover extra tokens.
Why it matters: Developers and enterprise users gain a more reliable, cost‑stable Claude model that flags uncertainty and writes higher‑quality code, while competitors face a stronger challenger. The upgrade also preserves pricing, meaning budgets stay flat even as performance rises, and the new dynamic workflow tool opens up tighter integration with software pipelines.
