Anthropic releases Claude Sonnet 5 — an agent-focused model cheaper than flagship Opus 4.8

Anthropic released Claude Sonnet 5 — an agent model cheaper than flagship Opus 4.8

On June 30, 2026, Anthropic announced the launch of Claude Sonnet 5 — the most agent-focused mid-tier model in the company's lineup, which on many tasks approaches the quality of the flagship Claude Opus 4.8 but costs less than half its price. The company said that Sonnet 5 is available starting today on all tariff plans.

What Claude Sonnet 5 can do

According to Anthropic, the model was created primarily to act rather than simply answer questions: the emphasis is placed on agent scenarios — executing multi-step tasks, using tools, working with code and files, and driving workflows to completion without constant human participation at each step. This distinguishes Sonnet 5 from previous generations of mid-tier models, which were primarily optimized for the quality of a single response rather than the reliability of executing long chains of actions.

• Launch date — June 30, 2026
• Claude Sonnet 5 is available immediately on all Anthropic plans, including the free one
• On many tasks, the model approaches the quality of flagship Claude Opus 4.8
• Sonnet 5 usage cost — less than half the price of Opus 4.8

How much Sonnet 5 costs compared to Opus 4.8

The key difference in the new model is the price-to-quality ratio. Anthropic positions Sonnet 5 as a model that covers the vast majority of practical tasks at nearly flagship level but costs companies launching AI agents at scale twice as much less. For businesses where token bills grow proportionally to the number of launched agents and processed tasks, such economics directly affect the decision to move from experiments to industrial deployment: agent scenarios typically require multiple calls to the model for each task — planning a step, calling a tool, checking the result — and the per-token price multiplies by this number of iterations.

Why agent models are becoming a separate class of products

The launch of Sonnet 5 continues a trend that all major AI labs are setting in 2026: instead of a single universal model, companies are releasing lineups divided by the price-to-speed-to-quality ratio, with a separate emphasis on autonomous task execution — from writing and editing code to working with browsers and internal company systems. Anthropic competes in this segment with agent products from OpenAI and Google, and the price per agent token becomes no less important a competition parameter than raw model quality on benchmarks.

What this means

The cost reduction of mid-tier agent models lowers the entry threshold for launching AI agents in production: companies no longer need to pay for a flagship model where Sonnet 5 quality is sufficient. This accelerates the race between labs for better price-to-quality ratio in the agent AI segment.

Frequently asked questions

How does Claude Sonnet 5 differ from Claude Opus 4.8?

Sonnet 5 is a mid-tier model that on many tasks approaches the quality of flagship Opus 4.8 but costs less than half its price; Opus 4.8 remains Anthropic's more powerful flagship model.

When did Claude Sonnet 5 become available?

Anthropic announced the launch on June 30, 2026, and on the same day opened access to the model on all company tariff plans.