Anthropic has launched Claude Sonnet 5.5 with over 30% faster performance, up to 30% lower task costs and a 1-million-token context window, targeting coding, research, document creation and professional workflows.
Anthropic has introduced Claude Sonnet 5.5, the second model in its Claude 5.5 family, positioning it as a faster, more economical option for coding, document creation, research, and other day-to-day professional work.
The release is interesting for a simple reason: Anthropic isn't trying to make Sonnet 5.5 merely another larger model.
Instead, the company is focusing heavily on how quickly and efficiently the model can finish a task.
Claude Sonnet 5.5 is a direct upgrade to Sonnet 5; it runs 30%+ faster and reduces task-related costs by up to 30% while keeping token-based pricing identical at $2 per million input tokens and $10 per million output tokens.
The model automatically adjusts its processing depth based on task complexity.
Users can manually scale this thinking effort up or down to manage processing times.
It is highly efficient for multi-agent software engineering workflows, outperforming flagship models like Opus 5.5 in continuous coding and bug-fixing loops.
It supports a 1-million-token context window, allowing users to process massive data sets, entire repositories, or long documents in a single prompt.
Loading post from https://x.com/claudeai/status/2104633115620823187…
It features an expanded single-output capacity of up to 128,000 tokens for extensive code generation and reporting.
It is the first Sonnet tier model deployed with top-tier cybersecurity safeguards to safely handle autonomous execution.
Claude Sonnet 5.5 delivers major performance gains over Claude Sonnet 5, which launched on June 30, 2026.
While both models share the identical baseline pricing of $2 per million input tokens and $10 per million output tokens, Sonnet 5.5 runs over 30% faster and reduces overall task costs by up to 30% due to its extreme token efficiency.
The new model delivers major benchmark gains, jumping from 10.3% in Sonnet 5 to 70.6% on Terminal Bench 4.0 for agentic coding and from 57% to 80.1% on OSWorld 2.1 for computer use.
Additionally, its real-world knowledge on GDPval-AA v2.1 jumps from 1,449 to 1,844, nearly matching Opus 5.5, prompting a shift from general safeguards to high-tier cybersecurity safeguards to handle its advanced capabilities.
| Claude Sonnet 5.5 vs Claude Sonnet 5 | |
|---|---|
| Specification | Details |
| Pricing | |
| Input Token Price | $2 per million tokens (identical for both) |
| Output Token Price | $10 per million tokens (identical for both) |
| Performance | |
| Speed Improvement | 30%+ faster than Sonnet 5 |
| Task Cost Reduction | Up to 30% |
| Context Window | 1 million tokens |
| Max Output Capacity | 128,000 tokens |
| Benchmarks | |
| Terminal Bench 4.0 (Sonnet 5) | 10.3% |
| Terminal Bench 4.0 (Sonnet 5.5) | 70.6% |
| OSWorld 2.1 (Sonnet 5) | 57% |
| OSWorld 2.1 (Sonnet 5.5) | 80.1% |
| GDPval-AA v2.1 (Sonnet 5) | 1,449 |
| GDPval-AA v2.1 (Sonnet 5.5) | 1,844 |
Sonnet 5 often needed multiple back-and-forth turns to read or explore directories before executing code.
Sonnet 5.5 typically chains tools together effortlessly, gathering files, reading configurations and writing solutions in far fewer steps.
Unlike Sonnet 5, Sonnet 5.5 has fully integrated adaptive reasoning built natively into its architecture.
It dynamically adjusts its internal processing depth, letting users choose Low, Medium, or High processing effort levels to balance speed with deep thought.
The Native API is available directly through the Anthropic platform using the identifier Claude Sonnet 5.5.
It is also fully supported via Amazon Bedrock with regional data residency as well as through the Claude Platform on Amazon Web Services (AWS).
For enterprise deployments, the model is generally available on Google Cloud, while Microsoft Azure offers it for testing and production workloads through Microsoft Foundry.
Additionally, GitHub Copilot has broadly rolled out the model to developers under usage-based pricing for day-to-day feature building and bug fixing.
With Opus 5.5 already at the top of Anthropic's current lineup and Haiku 5.5 expected to follow, Sonnet 5.5 sits in the middle of a broader strategy.
It is intended to balance capability and operating cost, and judging from Anthropic's published results, to make that middle tier considerably more capable than the previous generation.
