Claude Sonnet 5

Claude Sonnet 5 is positioned as a next‑generation coding and reasoning model. Leaked benchmarks show over 80.9% on SWE‑Bench, vastly exceeding previous bests, and it can handle real‑world GitHub issue resolution, code generation, and refactoring with a 1‑million‑token context window that spans large codebases.

Compared to the former flagship Opus 4.5, Sonnet 5 delivers superior performance while rumored pricing sits around $2/$10 per million tokens (input/output) — a ~50% reduction. It is also faster and expected to enable multi‑agent workflows within Claude Code, essentially acting as a virtual dev team that boosts throughput dramatically.

All details come from early leaks, 404 error logs on partner platforms, and social‑media reports; Anthropic has not officially launched the model. Availability, final pricing, and capabilities may shift. Early access is limited to select enterprise customers, and the model may act as a stop‑gap until Fable 5 returns, leaving its long‑term support unclear.