Claude Sonnet 5 Arrives, Bringing Opus-Level Performance at Sonnet Pricing
Anthropic unveiled its new model, “Claude Sonnet 5.” Its standout feature is that it approaches the performance of the higher-tier Opus 4.8 on coding and autonomous task execution, at a lower cost. It’s available across all platforms, including Claude.ai, Claude Code, and the API.
Details
- Improved reasoning and planning: Strengthened ability to autonomously work through complex problems
- Tool use: Handles software operations like browser and terminal control more accurately
- Coding: Improved debugging and follow-through — testers noted that “tasks that would have stalled partway with previous Sonnet versions now get completed to the end”
- Safety: Compared to Sonnet 4.6, rates of hallucination and excessive sycophancy have decreased, along with improved refusal of malicious requests and resistance to prompt injection
Plans and pricing
| Plan | Details |
|---|---|
| Free / Pro | Provided as the default model |
| Max / Team / Enterprise | Offered as one of several selectable models |
| API | Available as claude-sonnet-5 |
Pricing is set at an introductory rate through August 31, 2026: $2 per million input tokens and $10 per million output tokens (standard pricing is $3/$15). Rate limits have also been raised across all platforms.
How to try it
- Claude.ai: Free/Pro plans automatically use Sonnet 5. Max/Team/Enterprise users can switch to it from the model selection menu
- API: Simply specify
model: "claude-sonnet-5"to start using it - Claude Code: After updating, Sonnet 5 is used with no additional configuration required