Claude Sonnet 5
Anthropic's mid-tier workhorse, released 30 June 2026 — a 1M-token context by default, a genuine agentic step up, and a launch discount that runs through August.
Claude Sonnet 5 specs at a glance
| Maker | Anthropic |
| Released | 30 June 2026 |
| Context window | 1,000,000 tokens (default and maximum) |
| Max output | Up to 128,000 tokens |
| API pricing (intro) | $2 / $10 per 1M tokens through 31 Aug 2026 |
| API pricing (standard) | $3 / $15 per 1M tokens from 1 Sep 2026 |
| Availability | Default model on Free & Pro · available on Max, Team, Enterprise |
| Best at | Agentic coding, high-volume production work, large-codebase tasks |
Calculate API costs for this model → Compare its monthly bill against 15 other rate cards.
What's new in Claude Sonnet 5
The headline change is behavioral. Early-access partners consistently describe Sonnet 5 as much more agentic than its predecessors: it finishes complex tasks where previous Sonnet models would stop short, and it checks its own output without being explicitly asked. For teams running Sonnet in agent loops, that reads as fewer babysitting turns per task.
The second change is context. Sonnet 5 supports the 1M-token context window by default — 1M is both the default and the maximum, with no smaller variant — plus 128K max output tokens, which makes whole-codebase and long-document work practical on the mid tier.
Pricing and the intro window
Sonnet 5 launched with introductory pricing of $2 per 1M input tokens and $10 per 1M output tokens, valid through 31 August 2026. After that, it moves to the standard Sonnet rate card of $3/$15. If you have a heavy batch workload, running it before September is a real ~33% saving.
Note that per-token comparisons against older Sonnets are complicated by tokenizer changes across Anthropic's 2026 lineup — benchmark your real cost per request, not just the rate card.
Who should use Claude Sonnet 5
- High-volume production workloads — near-Opus quality on coding and agentic work at roughly half Opus price (and less during the intro window).
- Agent builders — the self-checking, finish-the-task behavior is the model's signature improvement.
- Large-codebase and long-document work — 1M context standard, no premium tier needed.
- claude.ai users — it's already the default on Free and Pro; nothing to configure.
If you need the absolute frontier, look at Claude Opus 4.8 or Claude Fable 5; if cost is everything, Claude Haiku 4.5 or open-weight models like DeepSeek V4 undercut it.
How Claude Sonnet 5 compares
- Claude Sonnet 5 vs GPT-5.5 — full comparison →
- Claude Opus 4.8 overview →
- Claude Haiku 4.5 overview →
- DeepSeek V4 — the open-weights option →
Frequently asked questions
When was Claude Sonnet 5 released?
Anthropic released Claude Sonnet 5 on 30 June 2026. From day one it is the default model for Free and Pro plans on claude.ai.
How much does Claude Sonnet 5 cost?
Introductory pricing is $2 per million input tokens and $10 per million output tokens through 31 August 2026. After that it costs $3/$15 per million tokens.
What is Claude Sonnet 5's context window?
1 million tokens by default — there is no smaller context variant — with up to 128,000 output tokens.
What is Claude Sonnet 5 best at?
Agentic work. Early access partners found it finishes complex tasks where previous Sonnet models stopped short, and it checks its own output without being asked.
Is Claude Sonnet 5 free to use?
Yes — it is the default model for Free and Pro plans on claude.ai, and available to Max, Team and Enterprise users. API access is pay-per-token.