Claude Sonnet 5: Permanent Pricing, Agentic Upgrades, and Migration Advice
Claude Sonnet 5 launched as Anthropic’s coding and agent-focused Sonnet, and Anthropic made its $2/M input and $10/M output price permanent on August 10, 2026. It is a strong migration candidate for agentic production work, but tokenizer changes mean teams should compare tokens and total task cost—not just per-million rates.
By Todd · Published 2026-08-14 · Verified 2026-08-14
| Endpoint | claude-sonnet-5 |
|---|---|
| Input | $2/M |
| Output | $10/M |
| Released | 2026-06-30 |
| Permanent price | Confirmed 2026-08-10 |
| Verified | 2026-08-14 |
The permanent-price update
The commercial news is as important as the model launch: Anthropic made the $2/$10 Sonnet 5 price permanent on August 10. That gives production teams a more stable budget assumption than a temporary launch rate, while the exact model-page limits and availability should remain tied to Anthropic’s current documentation.
What changed from Sonnet 4.6
Anthropic positions Sonnet 5 for coding, agents, tool use, and knowledge work. The meaningful migration question is whether those capabilities reduce human intervention and retries enough to offset any change in tokenization or output length.
- Agentic coding and tool-use positioning
- Effort/reasoning controls should be validated in the target API surface
- Updated tokenizer may produce roughly 1.0–1.35× tokens depending on content
- Safety claims in provider materials remain attributed claims, not All AI Ask measurements
Migration economics
Keep Sonnet 4.6 as a control. Replay representative tasks, compare input and output token counts, tool-call loops, retries, and successful completion cost. The $2/$10 headline is useful, but it is not a complete unit-cost model when the tokenizer and agent trajectory change.
Who should use it—and who should wait
Use Sonnet 5 when reliable coding and agent orchestration are worth a managed Anthropic endpoint. Wait if your application is highly sensitive to exact token budgets, depends on undocumented effort settings, or has not yet validated tool schemas and safety review requirements.
Availability and test status
The release announcement and changelog are the source of the launch and pricing update. All AI Ask has not published a controlled first-party benchmark verdict for Sonnet 5, so this page makes no unsupported quality or speed ranking.
Explore the canonical data
Test it yourself
Run the same prompt across claude sonnet 5, grok 4.6, gemini 3.7 flash in one workspace.
Open the comparison playground →FAQ
What is Claude Sonnet 5’s permanent price?
Anthropic made the launch price permanent at $2 per million input tokens and $10 per million output tokens on August 10, 2026.
Can Sonnet 5 cost more than the headline rate suggests?
Yes. Anthropic reports tokenizer changes can produce roughly 1.0–1.35× tokens depending on content, so measure total task tokens and retries.