Claude Sonnet 5.5

• 🆕 Anthropic's Sonnet-class upgrade, released September 2026, succeeding Sonnet 5.
• 📏 1M-token context window with up to 128K output tokens.
• 🧠 Adaptive thinking with selectable effort levels for depth-versus-latency tradeoffs.
• 🔧 Tuned for well-scoped work: features, bug fixes, documents, slides, spreadsheets.
• 👁️ Vision and document understanding, function calling, and web search.
• ⚡ Positioned as the faster, lower-cost complement to Opus 5.5.
• 💬 Anthropic says it writes and communicates more clearly than Sonnet 5.
• 🔒 Behavioral audit: matches or improves on Sonnet 5 across most measures.

Anthropic

Anthropic PBC is an American artificial intelligence company headquartered in San Francisco. Structured as a public benefit corporation, the lab develops large language models under the Claude name, with a research emphasis on building reliable, steerable,…

Claude Sonnet 5.5 is Anthropic's mid-tier Claude model, released in September 2026 as the direct successor to Claude Sonnet 5. In Anthropic's own framing, it is the faster, lower-cost complement to Claude Opus 5.5: where the Opus tier targets complex work requiring careful judgment, Sonnet 5.5 is aimed at well-scoped everyday tasks such as building features, fixing bugs, and producing polished documents, slides, and spreadsheets.

Structurally it carries forward the Sonnet 5 platform profile — a 1M-token context window that is both default and maximum, 128K maximum output tokens, adaptive thinking, and text, image, and file inputs. Capability-wise the catalog lists reasoning, vision, code optimization, function calling, and web search, so integrators moving from Sonnet 5 largely keep the same tool and feature surface. Anthropic's documentation now describes Sonnet 5.5 as the current Sonnet model and recommends migrating from Sonnet 5 for improved performance.

On generational gains, Anthropic reports that on Terminal-Bench 4.0 — a benchmark for multi-step professional tasks in a command line — Sonnet 5.5 at Medium effort (the default in the Claude apps) exceeds Sonnet 5's best score at a small fraction of the cost per task. Anthropic also says the model communicates and writes more clearly than its predecessor, a change echoed in platform listings.

On safety, Anthropic's automated behavioral audit of roughly 1,850 scenarios found Sonnet 5.5 improves on or matches Sonnet 5 on most measures of alignment, misuse resistance, and honesty, and performs strongly on newer containment evaluations, with Opus 5.5 still slightly ahead overall.