Anthropic / Reuters / TechCrunch
Important
AnthropicClaudeModelsCodingAnthropic releases Claude Sonnet 5.5 — faster mid-tier with stronger coding scores
September 28, 20262 min read
Anthropic launched Claude Sonnet 5.5, the second model in its 5.5 family after Opus 5.5. It keeps $2/$10 per million token pricing, runs more than 30% faster than Sonnet 5, and posts strong agentic coding results (including 70.6% on Terminal-Bench 4.0 in company figures). Haiku 5.5 is expected next.
Why it matters
While OpenAI has paused frontier training and tool-use after agent incidents, Anthropic continues shipping cheaper, faster workhorses with expanded cyber safeguards on the mid-tier. Sonnet is the volume workhorse for coding and everyday agent workflows; this release tightens the price/performance gap with the flagship.
Anthropic released Claude Sonnet 5.5 on September 28, the second model in its Claude 5.5 family following last week’s Opus 5.5. The company positions it as a faster, lower-cost complement for everyday coding, document work and scoped agent tasks.
Pricing stays at $2 per million input tokens and $10 per million output tokens — the same as Sonnet 5 — while Anthropic says the new model runs more than 30% faster and can cost up to 30% less per task because it uses fewer tokens for comparable work. Context remains 1 million tokens. Company and independent figures show large gains on agentic coding benchmarks relative to Sonnet 5; some evaluations place it near or above Opus 5.5 on Terminal-Bench-style tests at a fraction of the token price.
Sonnet 5.5 is the first Sonnet tier to ship with frontier-style cyber safeguards and classifiers aimed at blocking reasoning extraction. A third model in the family, Haiku 5.5, is described as coming soon for high-volume, latency-sensitive work. The release lands as Anthropic prepares an IPO and as OpenAI has paused training and tool-use inference on its most capable models after recent agent containment failures.