News
Claude Sonnet 5.5 Arrives: 30% Faster, Same Price, and Takes Aim at GPT-6 Sol
Anthropic has released the second model in the Claude 5.5 series, Sonnet 5.5. It responds over 30% faster than its predecessor, keeps API prices unchanged, and jumped from 10.3% to 70.6% on the Terminal-Bench 4.0 coding benchmark. Anthropic claims it 'clearly outperforms' OpenAI's GPT-6 Sol on long-horizon knowledge work.
On September 28, Anthropic released Claude Sonnet 5.5 — the second model in the Claude 5.5 series. The first was Opus 5.5, the flagship for maximum intelligence. Sonnet 5.5 sits one tier below: the everyday workhorse. Coding, bug fixes, documentation, slides and spreadsheets — Anthropic says it is optimized for these “clearly bounded everyday tasks.”
Key numbers
- Responses are more than 30% faster than Sonnet 5.
- Most tasks use fewer tokens, cutting real-world costs by up to 30%.
- API list prices are unchanged: $2 per million input tokens, $10 per million output tokens, $0.20 per million tokens for cache reads.
- Model name
claude-sonnet-5-5, already live on the Claude platform, AWS, Google Cloud and Azure, with zero data retention.
The speed and price figures deserve a second look. Prices didn’t rise while speed did — more for the same money. OpenAI’s GPT-6 Sol, at the same $2/$10 price point, now has a direct competitor.
Coding scores jumped sharply
The hardest data point in the release comes from Terminal-Bench 4.0, a benchmark of agentic coding ability. Sonnet 5.5 scored 70.6%; the previous-generation Sonnet 5 managed only 10.3%.
Terminal-Bench 4.0 isn’t a question set — it has models complete programming tasks end to end in a real terminal environment. A jump this large shows Sonnet 5.5 has shifted into a different gear on long multi-step tasks.
Naming GPT-6 Sol directly
Anthropic isn’t being shy this time. It says that on GDPval-AA — a benchmark covering tasks across 44 occupations — Sonnet 5.5 scores close to its own Opus 5.5.
More pointedly: “On long-horizon knowledge work, Sonnet 5.5 clearly outperforms OpenAI’s GPT-6 Sol.”
There’s fire in that line. GPT-6 Sol is OpenAI’s mid-tier model, which cut prices by 50% on September 22 with a value pitch. Anthropic putting a rival’s name directly in its launch materials sends a clear message: it wants the mid-tier market too.
One caveat: these are Anthropic’s own benchmark results, not yet fully verified by third parties. Vendor self-tests tend to run optimistic — discount accordingly.
Who it’s for
If you’re on Sonnet 5, upgrading is a near no-brainer: same price, faster, and cheaper on many tasks.
If you’re torn between Sonnet 5.5 and GPT-6 Sol, start with your workflow. For tasks spanning terminals, long steps and multi-tool calls, Sonnet 5.5’s benchmark numbers are stronger; OpenAI cut prices harder — GPT-6 Sol and Luna both dropped 50% on September 22. OpenAI is more aggressive on price; Anthropic has answered on capability.
Watch for third-party testing over the coming days. Real-world speed and token consumption are what will decide whether this model is really worth it.