claude-sonnet-5 — pricing, context window and limits
| Status | current |
|---|---|
| Input / MTok | $3 |
| Output / MTok | $15 |
| Context window | 1,000K tokens |
| Max output | 128K tokens |
| Min cacheable prefix | 1,024 tokens |
| Effort levels | low, medium, high, xhigh, max |
| Thinking | adaptive-default |
| Platforms | Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry |
Near-Opus quality at Sonnet pricing
Sonnet 5 is built to close the gap between Sonnet and Opus specifically on coding and agentic work, at Sonnet's own price point rather than Opus's — and it's the first Sonnet generation to carry the second-highest effort tier at all, a capability previously exclusive to the Opus line.
Two migration traps worth checking before you move real traffic
The tokenizer here counts noticeably more tokens than the immediately prior Sonnet generation for
identical text, at an unchanged sticker price — a real cost consideration invisible if you're only
looking at the per-token rate. And thinking runs by default now, which it didn't on the prior
generation, meaning max_tokens has to account for thinking overhead it previously didn't need
to. See what actually changed from Sonnet 4.6 to Sonnet 5
for both in full.
A temporary introductory rate is active
This model currently carries a promotional per-token rate below its standard price, active for a limited window. See what changes when an intro price expires for what to budget for once that window closes.
Related
See migrating: Sonnet 4.6 to Sonnet 5 and Sonnet 5 vs Haiku 4.5 for how it stacks up against the lineup's cheaper option.
Verified 2026-08-08 against claude-api skill — shared/models.md + SKILL.md model table (canonical: https://platform.claude.com/docs/en/about-claude/models/overview, https://platform.claude.com/docs/en/pricing).