Anthropic makes Claude Sonnet 5 the default for Free and Pro users

The model is available across Claude Code and major cloud platforms, with introductory API pricing in place until August 31.

Anthropic has made Claude Sonnet 5 the default model for Claude Free and Pro users, bringing updated coding, tool use and autonomous agent capabilities to its widely used mid-range model.

Claude Sonnet 5 became available on June 30, 2026 across all Claude plans. Max, Team and Enterprise customers can select the model, while developers can access it through Claude Code, the Claude API, Amazon Bedrock, Google Cloud Vertex AI and Microsoft Foundry.

Developers can use the model through the API under the name claude-sonnet-5. Anthropic has introduced temporary pricing of $2 per million input tokens and $10 per million output tokens until August 31, 2026.

From September 1, Claude Sonnet 5 will cost $3 per million input tokens and $15 per million output tokens, matching the standard price of its predecessor, Claude Sonnet 4.6.

Anthropic said the model improves on Sonnet 4.6 in coding, reasoning, tool use and knowledge work, while narrowing the capability gap with the more expensive Claude Opus 4.8. It also has a one million-token context window and can generate outputs of up to 128,000 tokens.

Sonnet 5 positioned as a drop-in upgrade

Anthropic has designed Claude Sonnet 5 as a direct replacement for Sonnet 4.6, allowing developers to change the model identifier without rebuilding their existing API integrations.

Rahul Patil, chief technology officer at Anthropic, wrote on LinkedIn:

"It's also a drop-in upgrade. Same API, same price, same speed targets as Sonnet 4.6. Swap the model string and you get better results immediately on multi-file code changes, long agent runs, and browser automation, with a 1M token context window and 128K output."

Anthropic said Sonnet 5 can plan tasks, use tools including browsers and terminals, and operate autonomously during longer workflows. The model is intended for recurring production tasks that require more capability than Anthropic's smaller Haiku models but do not require the full performance of its Opus range.

The developer said Sonnet 5 could complete complex tasks that earlier Sonnet models stopped before finishing and could check its own work without being explicitly instructed to do so. Those findings were based on feedback from early access partners and have not been independently verified in the information provided.

Claude Sonnet 5 also allows users to adjust the model's effort level, giving developers control over the balance between cost, token use and performance for different projects.

Anthropic has increased rate limits across Claude Chat, Cowork, Claude Code and the Claude Platform to support the higher token use associated with more demanding effort settings.

The model uses an updated tokenizer, which changes how text is divided and billed. Anthropic said the same input may produce between 1 and 1.35 times as many tokens as it did with Sonnet 4.6, depending on the type of content.

Anthropic said the introductory pricing was intended to make the transition approximately cost-neutral despite the tokenizer change. Actual API costs will depend on the content processed and the effort settings selected.

Anthropic compares performance with Opus 4.8

Anthropic said Claude Sonnet 5 delivers its largest capability improvement between two Sonnet generations and can match Claude Opus 4.8 on some agentic tasks when operating at higher effort levels.

The comparisons included Anthropic's tests of agentic web search through BrowseComp and computer use through OSWorld-Verified. Anthropic said Sonnet 5 provided more cost and performance options than Sonnet 4.6, particularly at medium effort levels.

Patil wrote:

"Over the past year, the clearest capability gains have come from our largest models. This is the biggest jump we’ve ever seen for a Sonnet model: roughly Opus 4.7 class on coding and agent tasks. For builders deciding what to run in production, it's the best cost to capability ratio in the Claude lineup."

Anthropic's official announcement compared Sonnet 5 more directly with Opus 4.8, stating that its overall performance was close to the larger model while remaining less expensive.

The comparison does not mean Sonnet 5 replaces Opus 4.8 across all workloads. Anthropic continued to position Opus for its most difficult tasks, Sonnet for high-volume daily work and Haiku for applications where speed and throughput are the primary requirements.

Anthropic also corrected a BrowseComp chart on June 30 after finding that the original version used a simpler methodology than its standard agentic search evaluation. The revised chart used a 10 million-token budget with compaction and programmatic tool calling, which Anthropic said increased the reported performance of Sonnet 5.

Safety tests find improvements and remaining gaps

Anthropic said its pre-deployment assessments found lower rates of hallucination, sycophancy and undesirable agent behavior in Claude Sonnet 5 than in Sonnet 4.6.

The model was also more likely to refuse malicious requests and resist prompt injection attempts, according to Anthropic's evaluations.

However, Sonnet 5 recorded higher rates of misaligned behavior than Anthropic's Claude Opus 4.8 and Claude Mythos Preview in the developer's automated behavioral audit.

Anthropic said Sonnet 5 had substantially weaker cybersecurity capabilities than its current Opus models. In an evaluation involving patched vulnerabilities in Firefox 147, Sonnet 5 did not produce a complete working exploit, although it recorded a slightly higher rate of partial success than Sonnet 4.6.

Anthropic has enabled real-time cybersecurity safeguards by default for Sonnet 5. The model is also included in Anthropic's Cyber Verification Program on the native Claude Platform, Amazon Web Services and Microsoft Foundry, with support for Google Cloud Vertex AI due to follow.

The introductory API price will remain available through August 31, before the higher standard rates take effect on September 1. Claude users and developers can access Sonnet 5 now through Anthropic's products and supported cloud platforms.