Claude (Sonnet 5)

What it is

Claude is Anthropic's model family, powering claude.ai and the Claude API used by many third-party tools. Anthropic trains Claude using Constitutional AI — instead of relying purely on human raters during RLHF, the model is also trained against a written set of principles it critiques its own outputs against, aimed at making its behavior more consistent and easier to reason about.

The current lineup

As of this writing, Anthropic's active lineup is the Claude 5 family:

• Claude Sonnet 5 (released June 30, 2026) — the balanced, general-purpose model, shipping with a 1M-token context window at introductory pricing.
• Claude Opus 4.8 — Anthropic's highest-capability model, leading independent aggregate benchmarks like the Artificial Analysis Intelligence Index (61.4) and topping the active SWE-bench Pro leaderboard at 69.2%.
• Claude Fable 5 (released July 1, 2026) — a "Mythos-class" model purpose-built for long-horizon agentic coding work, reported at 80.3% on SWE-bench Pro in third-party testing.
• Claude Haiku 4.5 — the fast, low-cost tier of the same generation.

Where these numbers come from

Anthropic's own release announcements for Claude Sonnet 5, Opus 4.5/4.8, and the surrounding model cards are the primary sources for this entry. Given how quickly this line moves — Anthropic has shipped multiple named releases within the same year — treat the specific benchmark figures above as a snapshot, and update this entry via the admin dashboard whenever Anthropic publishes a newer model card.

Notable traits

• Very large context windows (1M tokens on Sonnet 5) among the largest offered by any closed flagship model.
• Anthropic publishes a detailed "system card" with each release covering safety evaluations in more depth than most competitors.
• Closed weights — available through Anthropic's API, claude.ai, and major cloud marketplaces.