Use Claude Sonnet 5 for cheaper agent workflows before moving to Opus

Confirmed: Anthropic launched Claude Sonnet 5 on June 30, 2026 as its new Sonnet-class model for agentic coding, tool use, browser and terminal work, and everyday professional tasks. The practical answer is simple: teams can now test a lower-cost Sonnet model before defaulting every long-running agent job to Opus-class pricing.

What changed

Anthropic says Claude Sonnet 5 can plan, use tools such as browsers and terminals, and run autonomously at a level that recently required larger and more expensive models. The company positions it as a clear upgrade over Claude Sonnet 4.6 for reasoning, tool use, coding, and knowledge work.

The model is available from today across Claude plans. Anthropic says it is the default model for Free and Pro users, available to Max, Team, and Enterprise users, available in Claude Code, and available on the Claude Platform as claude-sonnet-5.

The API launch includes introductory pricing of $2 per million input tokens and $10 per million output tokens through August 31, 2026. After that, Anthropic lists standard pricing at $3 per million input tokens and $15 per million output tokens.

Why this is early

This is early but confirmed. Anthropic's own launch post is live, the Claude Platform release notes list the claude-sonnet-5 model ID, and the Claude Code changelog says Sonnet 5 is now the default model in Claude Code after updating to version 2.1.197.

Independent coverage has also started. TechCrunch frames the release as a cheaper way to run agents and notes the broader market shift: agentic capability is becoming a baseline expectation across model tiers, while cost and reliability are becoming the real decision points.

Key takeaways

• Claude Sonnet 5 is a confirmed Anthropic model release for agentic coding, tool use, browser/terminal work, and professional knowledge tasks.
• It is available in Claude Chat plans, Claude Code, and the Claude Platform API under claude-sonnet-5.
• Introductory API pricing runs through August 31, 2026 at $2 input / $10 output per million tokens.
• Anthropic says Sonnet 5 is close to Opus 4.8 on some agentic workflows, while Opus 4.8 remains the higher-accuracy choice for harder tasks.
• Migration needs testing because the Claude Platform notes list behavior changes, including adaptive thinking defaults, removed manual extended thinking, sampling parameter restrictions, and a new tokenizer.

Availability and access

Users can try Sonnet 5 now where Anthropic has enabled it. Anthropic says Free and Pro plans get it as the default model, while Max, Team, and Enterprise users can access it as well. Claude Code support is tied to the current changelog entry, which says to update to version 2.1.197.

For developers, the key API identifier is claude-sonnet-5. The Claude Platform release notes say Sonnet 5 supports a 1M token context window, 128k max output tokens, and the same set of tools and platform features as Sonnet 4.6 except Priority Tier.

There are migration details worth checking before a direct swap. The release notes say adaptive thinking is now on by default, manual extended thinking is removed, and non-default sampling parameters return a 400 error. Anthropic also says the updated tokenizer can produce roughly 1.0x to 1.35x as many tokens depending on content.

Practical LinkLoot angle

The practical workflow is to route more routine agent work to Sonnet 5, then reserve Opus 4.8 for jobs where accuracy matters more than marginal cost. That can include coding agents, browser automation, research cleanup, CRM updates, and multi-step internal operations where "finishes the task" matters more than peak benchmark bragging rights.

For LinkLoot readers building AI automations, the release is less about a new chatbot and more about the cost-performance layer under agents. If a workflow was too expensive on Opus but too brittle on older Sonnet models, Sonnet 5 is the obvious retest candidate.

Use LinkLoot's AI workflow automation guide as the next checklist layer: identify the repetitive work, measure token cost, add review gates, and only then decide whether Sonnet 5 belongs in the default route.

What to verify before you act

• Confirm claude-sonnet-5 appears in your Claude Console, model picker, or provider surface before changing production routes.
• Re-run token-count checks because the tokenizer can increase token counts for the same input.
• Test API requests that set temperature, top_p, or top_k; the release notes say non-default values return a 400 error on Sonnet 5.
• Check whether your workflow depends on Priority Tier, manual extended thinking, or older Sonnet 4.6 behavior.
• Review safety and compliance requirements for browser, terminal, and cyber-adjacent agent tasks before widening autonomous permissions.

Source check

Confirmed by Anthropic: Claude Sonnet 5 launched on June 30, 2026, is available across Claude plans, Claude Code, and the Claude Platform, and has introductory API pricing through August 31, 2026. Anthropic also states the model improves over Sonnet 4.6 on agentic reasoning, tool use, coding, and knowledge work.

Checked in Anthropic's developer docs: the Claude Platform release notes and Claude Code changelog both reflect the Sonnet 5 rollout. Those docs are useful for migration checks, but they are not treated as independent coverage.

Independent context: TechCrunch corroborates the release and frames Sonnet 5 as a lower-cost model for agents, while noting the competitive pressure from OpenAI and Google around agentic model capabilities.