Claude Sonnet 5.5 Ships at $2/$10 With 1M Context and Faster Agentic Coding

By AiCybr Editorial · Published September 30, 2026

Anthropic released Claude Sonnet 5.5 on September 28, 2026, with API pricing of $2 per million input tokens and $10 per million output tokens, a 1 million-token context window, and a 128,000-token maximum output. The model is available through the Claude API as claude-sonnet-5-5 and across Amazon Bedrock, Google Cloud, Microsoft Foundry, and Claude Platform on AWS.

Anthropic positions Sonnet 5.5 as the faster, lower-cost member of its current high-end lineup for everyday coding and knowledge work. The company reports output generation more than 30% faster than Sonnet 5 and says its internal testing found costs up to 30% lower per task, driven by lower token use while the base input/output token rates remain unchanged.

Coding is the largest reported benchmark gain. Anthropic reports 70.6% on Terminal-Bench 4.0, up from 10.3% for Sonnet 5 and above the 66.4% result it lists for Opus 5.5 at Xhigh effort. These are Anthropic's published evaluation results and should be compared at matched effort settings and evaluation configurations.

Claude Sonnet 5.5 specifications and pricing

The token rates match Sonnet 5. The economic change is task efficiency: Anthropic says Sonnet 5.5 typically completes equivalent work with fewer tokens. That distinction matters for agentic workloads, where repeated tool calls and long trajectories can dominate total token consumption.

Anthropic's launch material says Sonnet 5.5 batches tool calls more effectively and often finishes workflows in fewer steps. Several launch customers reported lower token use or faster completion in their own private evaluations. Those customer measurements use different workloads, so the resulting savings rates are workload-specific.

Coding benchmarks show a large jump over Sonnet 5

Anthropic's published launch table reports the following results:

Effort level is part of the performance equation. Claude Code and the Claude apps default Sonnet 5.5 to Medium effort, while the Claude Platform defaults to High. Anthropic's cost-versus-score charts show lower effort settings delivering much of the model's improvement at substantially lower task cost, while higher settings spend more tokens to pursue additional capability.

Anthropic also reports a GDPval-AA v2.1 score of 1844 for Sonnet 5.5 versus 1846 for Opus 5.5 and 1449 for Sonnet 5. Its launch notes say Artificial Analysis ran GDPval-AA and AA-Briefcase on a prerelease deployment affected by a structured-output bug that Anthropic subsequently fixed; the company expects any score impact to be small and downward.

API migration has one important thinking-setting change

Developers moving from Sonnet 5 can use claude-sonnet-5-5 on the Claude Platform. Anthropic specifically flags one migration requirement for applications that previously ran Sonnet with thinking disabled: those applications need to move to the new between_tools setting before switching to Sonnet 5.5. The setting keeps upfront thinking disabled while supporting the new model's reasoning behavior between tool calls.

The model supports zero data retention and is available across Anthropic's major cloud distribution channels. Platform-specific model identifiers differ: Amazon Bedrock uses anthropic.claude-sonnet-5-5, while the Claude API, Google Cloud, Microsoft Foundry, and Claude Platform on AWS list claude-sonnet-5-5.

New safeguards arrive with stronger cyber capability

Anthropic says Sonnet 5.5's cybersecurity capability is comparable to Claude Opus 5, making it the first Sonnet release to receive the company's higher-capability cyber safeguards and fallbacks. Higher-risk cybersecurity requests can fall back to Sonnet 5, while routine software development and defensive bug-fixing remain supported.

Sonnet 5.5 also introduces safety classifiers intended to resist large-scale reasoning-extraction and model-distillation attacks. Anthropic says its automated behavioral audit covered roughly 1,850 scenarios and found Sonnet 5.5 improved on or matched Sonnet 5 on most tested alignment, misuse-resistance, and honesty measures. These safety results are Anthropic's own evaluation findings; the accompanying system card provides the methodology and fuller results.

Deployment takeaways

For teams already using Sonnet 5, Sonnet 5.5 offers a direct migration path at the same list token rates, with the largest claimed gains concentrated in coding, tool use, output speed, and token efficiency. The 1M context window and 128K maximum output also make the model suitable for large repositories, long agent trajectories, and document-heavy workflows at Sonnet pricing.

The practical comparison with Opus 5.5 depends on workload and effort level. Sonnet 5.5 costs half as much per fresh input and output token, while both models list cache reads at $0.20 per million tokens. Anthropic continues to position Opus 5.5 for complex, open-ended work requiring sustained judgment and Sonnet 5.5 for faster, well-scoped work. Production buyers should benchmark completed-task quality, latency, token use, and total cost on their own agent traces before selecting a default model.

Sources

• Anthropic — Claude Sonnet 5.5 launch: https://www.anthropic.com/claude-sonnet-5-5
• Anthropic — Claude Sonnet 5.5 System Card: https://www.anthropic.com/claude-sonnet-5-5-system-card
• Claude Platform — Models overview and current specifications: https://platform.claude.com/docs/en/models/overview
• Reuters — Anthropic rolls out second Claude 5.5 model: https://www.reuters.com/technology/anthropic-rolls-out-second-claude-55-model-it-builds-toward-ipo-2026-09-28/