Anthropic launched Claude Sonnet 5 today, June 30, 2026. It is the most agentic Sonnet model yet, it is the new default for Free and Pro users, and it is priced to make running AI agents far cheaper. The release hit the front page of Hacker News within hours. Here is what actually shipped.

What launched

Sonnet 5 is the successor to Sonnet 4.6, and Anthropic built it to act rather than just answer. It plans, drives browsers and terminals, and runs autonomously for long stretches. That sort of sustained agentic work used to require larger, pricier models like Opus 4.8.

Key facts:

• API model string is claude-sonnet-5, codename Fennec.
• One million token context window, with text, image, and file inputs.
• Selectable reasoning effort: low, medium, high, max, and x-high.
• Available everywhere today: the Claude API, Claude Code, the Free and Pro default model, plus Amazon Bedrock, Microsoft Foundry, and Google Vertex.

The benchmarks

Sonnet 5 beats Sonnet 4.6 across the board and gets close to Opus 4.8 without overtaking it:

• SWE-bench Pro: 63.2 percent, versus 69.2 percent for Opus 4.8.
• OSWorld computer use: 81.2 percent, up from 78.5 percent for Sonnet 4.6.
• It posts a slight edge over Opus 4.8 on GPQA-AAA v2, a graduate-level reasoning test.

Opus 4.8 still wins the hardest coding and agentic tasks, especially at high reasoning effort. The fuller breakdown is in our Sonnet 5 vs Opus 4.8 comparison and the complete guide.

The price story

This is the headline for most teams. Sonnet 5 launches at an introductory $2 per million input tokens and $10 per million output tokens through August 31, 2026, then moves to $3 and $15. Opus 4.8 sits at $5 and $25.

Anthropic told reporters the goal of the introductory window is to let customers test Sonnet 5 against real workloads at the lowest possible cost during migration. There is a catch worth knowing: Sonnet 5 uses an updated tokenizer, so the same text can map to up to 1.35 times more tokens. The intro price is set to make the switch roughly cost-neutral rather than a flat discount. We unpack that in Sonnet 5 pricing explained.

Why this matters

For most of the past year, the clearest agentic gains came from Opus-class models that cost real money to run at scale. Sonnet 5 pulls a large chunk of that capability down into the mid tier. For teams running agents in production, where token spend compounds fast, a model that does most of what Opus 4.8 does at less than half the price changes the math.

It also arrives in a charged moment for Anthropic. The Mythos-class Fable 5 was banned by US export controls in June over its cyber capabilities. Sonnet 5 was deliberately not trained on cyber tasks and ships with lighter safeguards, so the risk of a similar pull is low. We cover that in Will the US government ban Sonnet 5?.

How to try it

• Set it as your model in Claude Code.
• Call claude-sonnet-5 through the API.
• Wire it into Aider or Cursor.

What early users report

Anthropic’s launch partners describe a consistent pattern: Sonnet 5 finishes work that older mid-tier models would leave half-done. Cursor reported agents that stay on plan and ship clean multi-step changes at an efficient cost. One team handed it a two-part task across two different systems and watched it complete both ends without stalling. Another described it investigating a bug, writing a test to reproduce it, fixing it, then confirming the fix by re-running the test, all in a single pass. For a model pitched at agentic work, that follow-through is the headline, more than any single benchmark.

How it fits the rest of the lineup

Sonnet 5 slots into Anthropic’s tiered family between the small, fast Haiku tier and the flagship Opus 4.8, with the restricted Mythos tier above that. What changed with this release is the distance between Sonnet and Opus. Previous Sonnet models trailed the flagship clearly; Sonnet 5 closes most of the gap and even edges Opus 4.8 on one graduate-level reasoning test. That is why the launch reframed the default model choice for so many teams overnight. For the head-to-head, see Sonnet 5 vs Opus 4.8, and for the generational view, Sonnet 5 vs Sonnet 4.6.

The catch worth repeating

The introductory price is real, but so is the tokenizer change. Sonnet 5 splits text differently, so the same prompt can use up to 1.35 times more tokens than on the old Sonnet. Anthropic set the intro rate to make the switch roughly cost-neutral, not a flat discount. If you are migrating, measure your real token usage rather than trusting the sticker rate. Our pricing explainer and token efficiency guide cover how to keep costs down.

Frequently asked questions

When was Claude Sonnet 5 released? June 30, 2026.

Is Sonnet 5 available on the free plan? Yes. It is the default model for both Free and Pro users.

How much cheaper is Sonnet 5 than Opus 4.8? At introductory pricing, Sonnet 5 input is $2 versus $5 for Opus 4.8, and output is $10 versus $25. Factor in the new tokenizer when estimating real costs.

Does Sonnet 5 support computer use? Yes. It scored 81.2 percent on OSWorld and is designed to drive browsers and terminals.

The bottom line

Sonnet 5 is the new value pick for agentic work. It is not the most capable Claude, but it is close enough for most jobs and far cheaper. With introductory pricing running through August, it is worth testing against your real workloads now. For the deep dive, start with the complete guide.