Claude Sonnet 5 Is Out: Near Opus Performance at One-Third the Cost
Anthropic launched Claude Sonnet 5 on June 30, its most capable mid-tier model to date. It scores 63.2% on SWE-Bench Pro, nearly matching Opus 4.8, at an introductory price of $2 per million input tokens.
Anthropic shipped Claude Sonnet 5 on June 30, and it collapses the performance gap between its mid-tier and flagship models in ways that matter for production workloads.
Benchmark scores
The headline number: 63.2% on SWE-Bench Pro, compared to 58.1% for Sonnet 4.6 and 69.2% for Opus 4.8. That puts Sonnet 5 most of the way from its predecessor to the flagship — at a fraction of the price.
Other scores tell the same story. OSWorld-Verified (computer use) hits 81.2%, up from 78.5%. Terminal-Bench 2.1 jumps to 80.4% from Sonnet 4.6’s 67.0%. Humanity’s Last Exam with Tools lands at 57.4%, within half a point of Opus 4.8’s 57.9%.
What changed
Anthropic positions Sonnet 5 as its primary model for agentic pipelines — the kind that chain multiple tool calls, run inside Claude Code, operate browsers, and execute long multi-step tasks autonomously. The company describes it as capable of “a level that just a few months ago required larger and more expensive models.”
SWE-Bench Verified reaches 85.2%. BrowseComp (web research accuracy) lands at 84.7% in single-agent mode and 86.6% in multi-agent. FrontierCode v1 comes in at 38.8%.
The model also ships with an updated tokenizer — the same kind Anthropic introduced with Opus 4.7 — which means the same text can now map to roughly 1.0 to 1.35 times more tokens depending on content type. Worth adjusting cost estimates if you’re processing code-heavy workloads.
Pricing and availability
Introductory pricing runs $2 per million input tokens and $10 per million output tokens through August 31, 2026. After that, it moves to standard pricing: $3 input / $15 output.
Sonnet 5 is available immediately across all Claude plans. It replaces Sonnet 4.6 as the default model for Free and Pro tiers. Max, Team, and Enterprise subscribers also have access from day one. It’s live on the Claude API, Amazon Bedrock, and Microsoft Foundry, with Google Cloud Vertex AI coming shortly.
Why it matters
For developers running agents at scale, the cost calculus just shifted. Opus 4.8 at full flagship pricing made sense only for tasks genuinely requiring the top model. Sonnet 5 at $2/$10 introductory — delivering nearly identical scores on agentic benchmarks — makes Opus the edge case rather than the default.
Anthropic is also opening a credits program: up to $30,000 in use credits for research projects that span domains, with an early focus on biology and biomedical science. Applications are open through July 15.
For most production agent pipelines, Sonnet 5 just became the obvious choice.