Anthropic Launches Claude Sonnet 5.5 With 30% Faster Responses at Sonnet 5 Pricing
Anthropic released Claude Sonnet 5.5 this week, an updated version of its mid-tier model that the company says responds more than 30% faster than Claude Sonnet 5 while keeping the exact same API pricing: $2 per million input tokens and $10 per million output tokens. The speed gain comes primarily from improved tool-call batching, which lets the model group multiple actions into fewer round trips when it is working through multi-step agentic tasks. Anthropic began rolling the model out across the Claude API, Claude Code, and claude.ai this week, with availability on Amazon Bedrock and Google Cloud Vertex AI following shortly after.
What Happened
Table of Contents
Claude Sonnet 5, which Anthropic launched on June 30, 2026, was already pitched as the company’s “most agentic mid-tier model yet,” running close to flagship Opus 4.8 on many tasks at under half the cost. Sonnet 5.5 keeps that positioning but targets the latency and throughput side of the equation instead of raw reasoning quality. Anthropic engineers said the core change is in how the model plans sequences of tool calls: rather than issuing one function call, waiting for a result, and issuing the next, Sonnet 5.5 batches related calls — for example, several file reads or several API lookups — into a single round trip whenever the calls do not depend on each other’s output.
Anthropic frames this as a cost improvement as much as a speed one: because agentic workflows are billed by total tokens processed across a task, and because Sonnet 5.5 needs fewer orchestration turns to complete the same job, internal task-completion benchmarks measuring end-to-end cost per completed coding or research task dropped alongside the latency gains, even though the per-token price is unchanged. The company said the model retains full backward compatibility with existing Sonnet 5 API calls and prompts, meaning most integrations can switch model IDs without code changes.
Why It Matters
Speed has become a competitive front in its own right among frontier AI labs, alongside raw benchmark scores. As agentic coding and research tools chain together dozens or hundreds of model calls per task, latency compounds: a model that is 30% faster per call can finish a long agent run in a meaningfully shorter wall-clock time, which matters directly to developers paying for compute-heavy CI pipelines, customer-support automation, and autonomous coding agents that run unattended for extended sessions.
Holding pricing flat also matters strategically. Rather than compete purely on lowering headline token prices — a race that squeezes margins across the industry — Anthropic is competing on getting more completed work out of the same dollar. For enterprise customers running Claude inside coding agents, customer service systems, or internal research tools, that framing changes how they evaluate the upgrade: it is a “free” performance improvement on infrastructure they have already budgeted for, since existing Sonnet 5 pricing agreements carry over.
How Sonnet 5.5 Compares
Sonnet 5.5 sits in the same tier of Anthropic’s lineup as Sonnet 5, below flagship Opus 4.8 and above the smaller, cheaper Haiku line. Key specs and changes include:
- Pricing: $2 per million input tokens, $10 per million output tokens — identical to Claude Sonnet 5
- Speed: Anthropic reports more than 30% faster response times on agentic, multi-tool-call workloads
- Cost per completed task: lower in internal benchmarks due to fewer orchestration round trips, not a lower token rate
- Compatibility: drop-in replacement for existing Sonnet 5 integrations via the Claude API
- Availability: Claude API, Claude Code, and claude.ai at launch; Amazon Bedrock and Google Cloud Vertex AI to follow
Anthropic has not published a fully updated benchmark suite alongside the release, focusing its announcement narrowly on latency and tool-call efficiency rather than claiming broad reasoning gains over Sonnet 5. That is a deliberate contrast with Opus-tier releases, which the company typically ties to state-of-the-art claims on coding and reasoning benchmarks.
What Developers Get
For teams already building on Sonnet 5, the practical upgrade path is straightforward: point existing API calls at the new model identifier and no other changes are required for most use cases, since prompt formatting, context window, and supported tool schemas remain the same. Anthropic said teams running high-volume agentic pipelines — code review bots, customer support triage, multi-step research agents — are likely to see the largest gains, since those workloads make the most tool calls per task and therefore benefit most from batching.
Frequently Asked Questions
Does Claude Sonnet 5.5 cost more than Sonnet 5?
No. Anthropic priced Sonnet 5.5 identically to Sonnet 5, at $2 per million input tokens and $10 per million output tokens. Any cost savings come from completing agentic tasks in fewer total tool-call round trips, not from a lower per-token rate.
How much faster is Sonnet 5.5 in practice?
Anthropic reports response times more than 30% faster than Sonnet 5 on agentic workloads involving multiple tool calls. Gains on simple, single-turn prompts without tool use are expected to be smaller, since the improvement is concentrated in tool-call batching.
Where can I access Claude Sonnet 5.5?
Claude Sonnet 5.5 is available now through the Claude API, Claude Code, and claude.ai. Anthropic said support for Amazon Bedrock and Google Cloud Vertex AI is coming shortly after the initial launch.
Related Articles
- Anthropic Says Its Claude Agents Discovered a New Enzyme System in Phage DNA
- Amazon Opens Its Seller Tools to Anthropic’s Claude in New AI Agent Push
Helpful Resources
- Anthropic — Official Claude model family page
- Anthropic Developer Docs — Claude API reference and pricing
