# Claude Sonnet 5 launches at $2 per million tokens

> Anthropic launched Claude Sonnet 5 on 30 June 2026 at an intro $2/$10 per million tokens.

*The brief: Anthropic's cheapest way to run AI agents lands at near-Opus quality — and it deliberately can't hack.*

By The FeaturedDaily Desk · FeaturedDaily
Canonical: https://featureddaily.com/news/claude-sonnet-5-launch-brief

> **Key:** **Claude Sonnet 5 is Anthropic's cheapest route to running AI agents** — near-Opus-4.8 quality at Sonnet prices, and it deliberately can't hack.

## What happened

Anthropic launched **Claude Sonnet 5** on **30 June 2026**, calling it its most agentic Sonnet yet — a model built to plan, use tools like browsers and terminals, and run autonomously at a level that months ago needed larger, pricier models. The API name is `claude-sonnet-5`. It is available on every plan and is now the **default model for Free and Pro** users, as well as Max, Team and Enterprise.

On performance, Anthropic says Sonnet 5 approaches Opus 4.8 for less money. It scored **63.2%** on an agentic coding benchmark (Opus 4.8: 69.2%; Sonnet 4.6: 58.1%) and slightly **outperforms** Opus 4.8 on a knowledge-work benchmark. An adjustable effort level lets developers trade cost against performance.

## The money

Pricing is the headline. Introductory rates run through **31 August 2026**, then step up to standard.

| Model | Price in/out (per M tokens) | Status |
| --- | --- | --- |
| Sonnet 5 (intro) | $2 / $10 | Out now |
| Sonnet 5 (from 1 Sep) | $3 / $15 | Scheduled |
| Opus 4.8 | $5 / $25 | Out |

> Sonnet 5 provides developers with lower-priced options of much higher quality than what was previously available, though Opus 4.8 remains the model of choice for higher accuracy.
> — [Anthropic](https://www.anthropic.com/news/claude-sonnet-5), 2026-06-30

> **Note:** **The catch:** Sonnet 5 uses a new tokenizer, so the same text can map to ~1.0–1.35× more tokens. Anthropic set the intro price so switching stays roughly cost-neutral versus Sonnet 4.6, but your real bill depends on your content.

## The safety angle

Sonnet 5 shows a lower overall rate of undesirable behaviours than Sonnet 4.6 and is safer in agentic use. Notably, Anthropic made an **explicit decision not to train it on cybersecurity tasks**, leaving it substantially weaker at exploit development than Opus 4.8 — a pointed choice in the same month the US pulled Fable 5 and Mythos 5 over cyber capability.

> TechCrunch frames Sonnet 5 as a cheaper way to run agents, but cautions that nearly all the concrete performance numbers sit inside Anthropic's own post, with partner testimonials drawn from early access rather than independent testing.
> — [TechCrunch](https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/), 2026-06-30

## Key takeaways

- Claude Sonnet 5 shipped on 30 June 2026 as Anthropic's most agentic Sonnet, built to plan and use tools autonomously.
- Intro pricing is $2/$10 per million input/output tokens through 31 August 2026, then $3/$15 — against Opus 4.8 at $5/$25.
- It approaches Opus 4.8 quality: 63.2% on an agentic coding benchmark, and it edges Opus on knowledge work.
- Anthropic chose not to train Sonnet 5 on cyber-offence, leaving it far weaker at exploit development than Opus 4.8.

## FAQ

### How much does Claude Sonnet 5 cost?
An introductory $2 per million input tokens and $10 per million output tokens through 31 August 2026, then $3/$15. For comparison, Opus 4.8 is $5/$25.

### Is Claude Sonnet 5 better than Opus 4.8?
It approaches Opus 4.8 rather than beating it: 63.2% versus 69.2% on an agentic coding benchmark, though it slightly outperforms Opus on a knowledge-work benchmark. Anthropic says Opus 4.8 remains the pick for higher accuracy.

### Can Claude Sonnet 5 be used for hacking?
Anthropic deliberately did not train it on cybersecurity offence, so it is substantially weaker at exploit development than Opus 4.8, while showing fewer undesirable behaviours overall.

### Are these benchmark numbers independently verified?
Not yet. Almost every figure comes from Anthropic's own launch post, and partner testimonials are early-access rather than independent. Independent benchmarking will settle real-world performance.

## Sources

- [Introducing Claude Sonnet 5](https://www.anthropic.com/news/claude-sonnet-5) — Anthropic, 2026-06-30
- [Anthropic launches Claude Sonnet 5 as a cheaper way to run agents](https://techcrunch.com/2026/06/30/anthropic-launches-claude-sonnet-5-as-a-cheaper-way-to-run-agents/) — TechCrunch, 2026-06-30
- [Anthropic Cuts AI Agent Costs With Claude Sonnet 5 Rollout](https://www.pymnts.com/news/artificial-intelligence/2026/anthropic-cuts-ai-agent-costs-with-claude-sonnet-5-rollout/) — PYMNTS, 2026-06-30