🤖 AI Curated
Anthropic unveiled Claude Sonnet 5 on June 30th, designating it as the default model for Free and Pro plans. It achieves 63.2% in agent coding, nearing Opus 4.8's 69.2%, and performs equally well in knowledge work. With introductory pricing of $2 for input and $10 for output (until 8/31), it's approximately 60% cheaper than Opus.
Anthropic unveiled its new model, 'Claude Sonnet 5,' on June 30th (local time). The core message is simple: it delivers performance close to the top-tier Opus 4.8 model, but at a much lower price. Anthropic introduced Sonnet 5 as "the most agent-like Sonnet ever," stating that its ability to autonomously use tools like browsers and terminals, plan, and execute tasks has reached a level that, just a few months ago, required larger, more expensive models.
The most notable change is its status as the 'default model.' From its launch, Sonnet 5 has been designated as the default model for Free and Pro plans, and it's also available to Max, Team, and Enterprise users. Developers can call it via API using the `claude-sonnet-5` identifier and use it directly within Claude Code. This means that without explicitly choosing a higher-tier model, the majority of users will automatically receive an upgraded baseline capability.
Performance varies by benchmark. Figures cited by various media outlets show that in agent coding, Sonnet 5 scores 63.2%, trailing Opus 4.8's 69.2% by about 6 points. However, in knowledge work evaluations, Sonnet 5 was found to be on par with or slightly ahead of Opus 4.8. In summary, the division of labor is clear: "Opus 4.8 for the most challenging coding and highest accuracy tasks, and Sonnet 5 for other everyday tasks."
The improvement over its predecessor is also significant. Sonnet 4.6, released in February 2026, scored 58.1% in agent coding, while Sonnet 5 achieved 63.2%, an increase of about 5 points. Anthropic explains that improvements have been made across the board in reasoning, tool use, coding, and knowledge work.
Pricing is the real weapon of this announcement. Sonnet 5 is available at an introductory special price until August 31st: $2 per million input tokens and $10 per million output tokens. After this period, the standard price will be $3 for input and $15 for output. Compared to the Opus series (around $5 for input and $25 for output), Sonnet 5 is about 40% cheaper at standard pricing and approximately 60% cheaper at introductory pricing, significantly reducing the cost burden for using it as a 'daily driver.'
In terms of safety, Anthropic stated that Sonnet 5 generally has a lower rate of undesirable behaviors than Sonnet 4.6 and is safer in agent environments. Specifically, its cybersecurity-related capabilities are designed to be lower than current Opus models, and cyber safety features are enabled by default upon deployment.
In summary, Sonnet 5 is less of a 'performance-disrupting new model' and more of a strategic move to redefine pricing and accessibility by bringing top-tier model capabilities to the default setting of mainstream plans. While Opus 4.8 remains for tasks requiring extreme accuracy, Sonnet 5 is positioned to cost-effectively handle most practical work below that level.
For domestic developers and creators, the perceived change is significant. As Sonnet 5 becomes the default for Free and Pro plans, the quality of agent-based automation improves without additional payment. With API introductory pricing being about 60% cheaper than Opus, the token cost burden for solo entrepreneurs and startups is reduced. However, since the price will rise to standard rates after August 31st, it's advisable for those planning large-scale calls to review their cost structure during the special offer period.
🤖 AI-curated from multiple sources. Verify accuracy with the originals (sources).