Anthropic released Claude Sonnet 5. Performance near Opus 4.8 across benchmarks but faster execution and materially lower cost structure. Pricing $2/$10 per million tokens (input/output) through Aug 31, then $3/$15. Opus 4.8 runs $5/$25 for reference.
Key feature set: autonomous task planning, browser/terminal access, self-correction without human intervention. 1M token context window standard. Positioning is clear—Opus 4.8 for edge-case complexity, Sonnet 5 for high-volume standard workloads (estimated 90% of use cases).
Anthropic also lifted API rate limits system-wide alongside launch.
Market question remains: where are the restricted frontier models (Fable 5, Mythos 5) rumored to be under US government export controls? No visibility yet.
Key feature set: autonomous task planning, browser/terminal access, self-correction without human intervention. 1M token context window standard. Positioning is clear—Opus 4.8 for edge-case complexity, Sonnet 5 for high-volume standard workloads (estimated 90% of use cases).
Anthropic also lifted API rate limits system-wide alongside launch.
Market question remains: where are the restricted frontier models (Fable 5, Mythos 5) rumored to be under US government export controls? No visibility yet.