Anthropic Launches Claude Sonnet 5 With Near-Opus Coding Performance
Anthropic's new Claude Sonnet 5 brings near-Opus 4.8 agentic coding performance at Sonnet pricing, with a 1M-token context window and introductory rates of $2/$10 per million tokens through August 31.
Image by Anthropic
Anthropic Launches Claude Sonnet 5 With Near-Opus Coding Performance#anthropic-launches-claude-sonnet-5-with-near-opus-coding-performance
Anthropic released Claude Sonnet 5 on June 30, positioning it as the most agentic Sonnet model the company has shipped. The model closes much of the performance gap with Opus 4.8 while launching at a lower price point, with introductory API pricing of $2 per million input tokens and $10 per million output tokens through August 31.
What Changed With Sonnet 5
Sonnet 5 is the first Sonnet-class model Anthropic is calling "Opus-adjacent" in agentic tasks. The model scores 63.2% on SWE-bench Pro, compared to 69.2% for Opus 4.8 and 58.1% for its predecessor, Sonnet 4.6. On knowledge-work benchmarks, Sonnet 5 slightly outperforms Opus 4.8. Anthropic says the model can make plans, drive browsers and terminals, and run autonomously through long, multi-step tasks at a level that, until recently, required larger and more expensive models.
The model ships with a native 1-million-token context window and is immediately available as the default model for Free and Pro plans on Claude.ai. Max, Team, and Enterprise users can access it alongside existing models. Claude Code users get Sonnet 5 as the new default, accessible via the claude-sonnet-5 API identifier.
Pricing and What It Means for Developers
The introductory price of $2 input / $10 output per million tokens through August 31 makes Sonnet 5 cheaper than Opus 4.8 ($5/$25), OpenAI's GPT-5.5, and Google's Gemini 3.1 Pro. Anthropic set the promotional rate to offset a tokenizer update that changes how input maps to tokens — roughly 1.0–1.35x more tokens for equivalent content. After August 31, the price reverts to standard Sonnet pricing of $3/$15 per million tokens.
For teams running agentic coding pipelines through Claude Code or the API, Sonnet 5 is now the default workhorse, with Opus 4.8 reserved for tasks that need higher accuracy on the hardest reasoning problems. The model is available on the Claude API, Amazon Bedrock (anthropic.claude-sonnet-5), Google Cloud Vertex AI, and Microsoft Foundry.
Safety Profile
Anthropic's pre-deployment evaluation found Sonnet 5 demonstrates a lower rate of undesirable behaviors than Sonnet 4.6, including better resistance to prompt injection attacks and lower hallucination and sycophancy rates. It is not at the safety level of Opus 4.8 or Mythos Preview. Notably, Anthropic found Sonnet 5 has substantially lower ability to perform dangerous cybersecurity tasks than the Opus models — the company shipped it with cyber safeguards enabled by default. No successful exploits were developed in Firefox vulnerability testing, compared to partial success rates on Opus models.
What's Unconfirmed
Anthropic has not disclosed user adoption data for the new model or a breakdown of how default-model switching will affect existing Pro subscribers mid-billing-cycle. The promotional pricing window ends August 31 without a stated commitment to extend it. Developers building against the API should factor in the tokenizer change when modeling per-task costs — the 1.0–1.35x token increase means the effective cost is not always lower than Sonnet 4.6 depending on input content type.
For developers evaluating the upgrade from Sonnet 4.6, the introductory window through August gives a practical period to benchmark costs and quality on real workloads before the standard $3/$15 price takes effect.
