Claude Sonnet 5 Release Date: When It Launched, Pricing, and What’s New
The Claude Sonnet 5 release date is June 30, 2026, when Anthropic shipped it as the successor to Sonnet 4.6. On that same day, Claude Sonnet 5 became the default model across Claude’s consumer plans, and Anthropic detailed the launch in its official announcement.
Positioned as the most agentic Sonnet yet, the model targets everyday professional work and coding at a lower price than Anthropic’s flagship. This guide covers the exact launch timing, pricing, availability, benchmarks, and what actually changed.
When Was Claude Sonnet 5 Released?
Anthropic released Claude Sonnet 5 on June 30, 2026, retiring Sonnet 4.6 as the standard Sonnet-class option. The model carries the API identifier claude-sonnet-5, which is what you reference when calling it programmatically.
The rollout was immediate rather than staged for consumers. On launch day, Sonnet 5 became the default model for Free and Pro users, and it was simultaneously offered to Max, Team, and Enterprise customers. The same model also shipped inside Claude Code and the Claude Platform, so developers could switch to it without waiting for a separate release.
Third-party platforms picked it up on day one as well. GitHub confirmed that Sonnet 5 became generally available in GitHub Copilot on June 30, 2026, running under Zero Data Retention and selectable in the model picker across VS Code, JetBrains, Xcode, and the Copilot CLI.
The naming also signals where the model sits in the lineup. Sonnet is Anthropic’s balanced tier, between the compact Haiku models and the flagship Opus line, so “Sonnet 5” tells you to expect strong general capability rather than the maximum ceiling of Opus 4.8. Replacing Sonnet 4.6 on the same day means there was no gap where the older model stayed the default.
Claude Sonnet 5 Pricing
Claude Sonnet 5 launched with introductory API pricing of $2 per million input tokens and $10 per million output tokens, in effect through August 31, 2026. After the introductory window, pricing settles at $3 per million input and $15 per million output, per Anthropic’s own pricing documentation.
That places it well below Anthropic’s flagship. Opus 4.8 costs $5 per million input and $25 per million output, so Sonnet 5 gives you a large share of frontier capability at a fraction of the per-token cost. You can confirm the current numbers on the Claude pricing page.
The table below summarizes how the introductory and standard rates compare with the flagship model.
| Model | Input (per 1M) | Output (per 1M) |
|---|---|---|
| Claude Sonnet 5 (intro, through Aug 31 2026) | $2 | $10 |
| Claude Sonnet 5 (standard) | $3 | $15 |
| Opus 4.8 | $5 | $25 |
On price and capability, Sonnet 5 reads as a direct, lower-cost alternative to OpenAI’s GPT-5.6 Sol and Google’s Gemini 3.5 Flash for routine agentic tasks.
Availability: Plans, Claude Code, API, and Integrations
Claude Sonnet 5 is broadly available rather than gated behind a premium tier. Because it became the default on Free and Pro at launch, most users interacting with Claude are already running Sonnet 5 without changing a setting.
Consumer and business plans
Every paid tier can use it. Free and Pro get it as the default, while Max, Team, and Enterprise customers can select it alongside other models. For anyone who wants to try the model in the browser without keys or setup, the extension at sonnet5.pro provides a quick path to chatting with Sonnet 5 directly.
The model is also the natural choice for high-volume work, since the introductory pricing makes long agentic runs and large context windows cheaper to operate than the flagship.
Developer surfaces and third-party tools
Developers reach Sonnet 5 through the Claude API using the claude-sonnet-5 model id, and inside Claude Code for terminal-based agentic workflows. Beyond Anthropic’s own surfaces, GitHub Copilot exposes it across the IDE and CLI, where early internal testing highlighted strong CLI-style performance and efficient prompt-cache utilization.
Benchmarks: Sonnet 5 vs Opus 4.8 vs Sonnet 4.6
The clearest way to place Sonnet 5 is against the model it replaces and the flagship it sits below. On agentic coding, Sonnet 5 scores 63.2%, a marked jump over Sonnet 4.6 at 58.1%, while Opus 4.8 still leads at 69.2%.
That gap is narrower than the price gap, which is the point of the release. Anthropic says Sonnet 5 approaches the performance of Opus 4.8, its most advanced widely available model, and it slightly surpasses Opus 4.8 on some knowledge-work benchmarks even though Opus retains the edge in pure agentic coding.
| Model | Agentic coding score |
|---|---|
| Opus 4.8 | 69.2% |
| Claude Sonnet 5 | 63.2% |
| Claude Sonnet 4.6 | 58.1% |
For teams choosing a default, the takeaway is that Sonnet 5 closes most of the quality distance to the flagship while costing far less to run at scale.
How Claude Sonnet 5 Compares to Rivals
The release was framed around cost as much as capability. Anthropic launched Sonnet 5 at a steep discount to its top model, a move widely read as a bid to win the high-volume agent market where per-token cost compounds quickly. Reporting on the launch described it as a cheaper way to run agents day to day, and that positioning is captured in TechCrunch’s coverage of the launch.
Against OpenAI and Google. Sonnet 5 is pitched as a direct, lower-cost alternative to OpenAI’s GPT-5.6 Sol and Google’s Gemini 3.5 Flash for everyday agentic tasks. It undercuts Opus 4.8, OpenAI’s GPT-5.5, and Google’s Gemini 3.1 Pro on price, while sitting slightly above Gemini 3.5 Flash. The pitch is not that it beats every frontier model, but that it delivers near-flagship results for the tasks most teams actually automate.
Against its own lineup. Within Anthropic’s own family, the choice is simpler than it looks. Opus 4.8 remains the pick when you need the absolute ceiling on agentic coding, and Sonnet 5 is the default for everything else, because the quality gap is small and the cost gap is large. That framing is why Sonnet 5, not Opus, is the model most users now run by default.
What’s New in Claude Sonnet 5
Sonnet 5 is described as the most agentic Sonnet model yet: it plans, reasons, uses tools such as a browser and terminal, coordinates subagents, and runs autonomously for extended tasks. That agentic framing is the headline change from Sonnet 4.6.
Two capabilities stand out beyond raw benchmarks:
- Self-correction. The model reviews its own output and fixes errors before you see the final answer, reducing the obvious mistakes that require a follow-up prompt.
- Improved safety. Anthropic reports stronger refusal behavior and better resistance to prompt-injection attacks in agentic settings.
On the security side, the launch shipped with a concrete result. Working with Mozilla, Anthropic tested the new Sonnet models against a Firefox 147 exploit benchmark, and both scored 0.0%; the underlying issue was patched in Firefox 148. Anthropic frames Sonnet 5 as carrying a lower dangerous-cyber risk profile than its most powerful systems, which is part of why it is comfortable making Sonnet 5 the broad default while keeping its most powerful capabilities on more tightly controlled models.
For everyday use, the practical upshot is fewer round-trips. A model that self-corrects, resists prompt injection, and runs tools autonomously handles multi-step tasks that previously needed a human in the loop, from browsing and summarizing to writing and testing code. That is the everyday-professional-work angle Anthropic emphasized at launch, rather than a single headline benchmark.
Our most agentic Sonnet yet, with top-tier intelligence for coding and everyday professional work.
Anthropic
If you want to verify the model id and start calling it, the setup is straightforward:
- Sign in to the Claude Platform and open the API keys section.
- Create a key with access to the model.
- Reference
claude-sonnet-5as the model in your request. - Send a small test prompt to confirm the response.
- Watch usage during the intro window to lock in the $2/$10 rates before September 2026.