Anthropic announced Claude Sonnet 4.6 as its most capable Sonnet model, featuring a 1M token context window in beta and upgraded abilities in coding, agent planning, computer use, and long‑context reasoning. The model is now the default on the company’s free and pro plans, with pricing unchanged from Sonnet 4.5 at $3/$15 per million tokens. Early adopters report significant gains in consistency, instruction following, and reduced hallucinations, often preferring Sonnet 4.6 over both Sonnet 4.5 and the earlier Opus 4.5. Benchmark results show strong performance across OSWorld, Vending‑Bench Arena, OfficeQA, and insurance computer‑use tasks, achieving 94% accuracy on the insurance benchmark. Safety evaluations indicate Sonnet 4.6 is as safe as or safer than recent Claude models, with improved resistance to prompt‑injection attacks. The model’s expanded context window allows it to hold entire codebases or lengthy documents in a single request, enhancing long‑horizon planning and complex code‑base work. Anthropic positions Sonnet 4.6 as a viable, cost‑effective alternative to Opus for users needing frontier‑level reasoning and agentic capabilities.
Introducing Claude Sonnet 4.6
Anthropic has released Claude Sonnet 4.6, its most capable Sonnet model to date, offering a 1M token context window in beta and upgraded skills in coding, agent planning, computer use, and long‑context reasoning. The model is the default on free and pro plans and costs $3/$15 per million tokens, the same as Sonnet 4.5. Early users report significant improvements in consistency, instruction following, and reduced hallucinations, often preferring Sonnet 4.6 over both Sonnet 4.5 and the earlier Opus 4.5. Benchmarks show strong performance on OSWorld, Vending‑Bench Arena, OfficeQA, and insurance computer‑use tasks, with a 94% accuracy on the insurance benchmark.
Why it matters
The release demonstrates that smaller, cost‑effective models can approach Opus‑level performance, potentially lowering barriers for enterprises to adopt advanced AI for coding, document analysis, and complex agentic workflows.