18 posts tagged with "models"

Claude Mythos Preview is Anthropic's most powerful model. Benchmarks, Project Glasswing access, pricing at $25/$125, why it's not public.

Sonnet 4.6 delivers Opus-level intelligence at $3/$15 per million tokens. 70% preferred over Sonnet 4.5 in coding. Full specs and benchmarks.

Claude Opus 4.6 leads Terminal-Bench 2.0, Humanity's Last Exam, GDPval-AA. Better coding, 1M context GA, 128K output, same $5/$25 pricing.

Opus 4.5 dropped to $5/$25 per million tokens with 76% fewer output tokens. The most efficient frontier model for Claude Code development.

Claude Haiku 4.5 launched October 2025 as the budget 4.5 model. Smart model switching made it the default for routine Claude Code tasks.

Claude Sonnet 4.5 launched September 2025 with top intelligence across most tasks, 1M beta context, and best-in-class coding performance.