Anthropic\u2019s Sonnet 5 promises near-Opus performance at lower cost, enabling cheaper autonomous workflows and forcing rivals to rethink agent pricing.<\/p>\n
In the context of growing competition for agent capabilities among mid-range models, Anthropic has released Claude Sonnet 5 \u2013 a more powerful and more agentic version of its mid-range personal model at a significantly lower price.<\/p>\n
It can plan, use tools like browsers and terminals, and act autonomously at a level that a few months ago would have required larger and more expensive models.<\/p>\n
\u2013 Anthropic<\/p>\n
Such statements align with claims from OpenAI and Google. OpenAI unveiled GPT-5.6 Sol, which is also considered the most agentive model the company has, with the ability to distribute tasks among plug-in agents for long-running autonomous processes. Google released Gemini 3.5 Flash in May, positioning it as a transition from a chat bot to an agent tool that plans, builds, and executes real work with minimal human involvement.<\/p>\n
Sonnet 5\u2019s Positioning and the Competitive Landscape<\/p>\n
The core idea behind Claude Sonnet 5 is that agentic capability has become a baseline expectation at any price level. The difference now is not who is better at agenting, but how cheaply and reliably this is done without constant oversight.<\/p>\n
At launch, Sonnet 5 promises performance close to Opus 4.8, but at a substantially lower price. Claude Sonnet 5 will become the default model for free and Pro plans and will be available with every subscription.<\/p>\n
From launch, the price will be $2 per million input tokens and $10 per million output tokens through August 31; after that it will jump to $3 per million input tokens and $10 per million output tokens. This combination makes Sonnet 5 cheaper than Opus 4.8, as well as GPT-5.5 and Gemini 3.1 Pro, though it still costs more than Gemini 3.5 Flash.<\/p>\n
The new model also shows substantial improvements over its predecessor Sonnet 4.6, released in February, in agentive metrics such as task understanding, tool use, coding, and other knowledge-based operations.<\/p>\n
Opus 4.8 remains the choice for higher accuracy in such tasks, but Sonnet 5 offers developers cheaper, yet higher-quality options than before.<\/p>\n
\u2013 Anthropic<\/p>\n
Between Sonnet 5 and Opus 4.8, users can adjust the level of effort to find the right balance between cost and performance.<\/p>\n
\u2013 Anthropic<\/p>\n
According to testers, Sonnet 5 also handles complex tasks well where previous versions stalled, and checks its own output without an explicit prompt.<\/p>\n
We tasked Claude Sonnet 5 with a two-step job \u2013 updating Salesforce account levels and sending a launch announcement to corporate contacts \u2013 and it completed the work from start to finish.<\/p>\n
\u2013 Daniel Shepard, Senior Engineer at Zapier<\/p>\n
Earlier this would stall halfway. For everyday automation this is the obvious choice.<\/p>\n
\u2013 Daniel Shepard, Senior Engineer at Zapier<\/p>\n
On safety, Sonnet 5 shows a lower level of unwanted behavioral actions, in particular less collaboration with malicious actors and scams, and better ability to refuse harmful requests and avoid prompt-injection attacks. However in some respects it still does not match Opus 4.8 and Claude Mythos Preview in alignment with safety.<\/p>\n
Claude Sonnet 5 refuses dangerous requests clearly and consistently, \u2013 according to Fabian Hedin, co-founder of Lovable.<\/p>\n
\u2013 Fabian Hedin, co-founder of Lovable<\/p>\n
\u201cAt Lovable we aim to provide powerful tools to millions of developers. A model that knows when to say no is just as important as one that can build,\u201d Hedin said.<\/p>\n
Overall, Claude Sonnet 5 confirms that agentic capability has become a baseline expectation in the modern market, and lowering costs while preserving quality could significantly reshape the competitive landscape among providers of agent-based solutions.<\/p>\n","protected":false},"excerpt":{"rendered":"Anthropic\u2019s Sonnet 5 promises near-Opus performance at lower cost, enabling cheaper autonomous workflows and forcing rivals to rethink…\n","protected":false},"author":2,"featured_media":91031,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[47125,1365,53,3154,38227,182,48047,48046,66,31869],"class_list":["post-91030","post","type-post","status-publish","format-standard","has-post-thumbnail","category-anthropic","tag-agent-models","tag-ai-automation","tag-anthropic","tag-anthropic-claude","tag-anthropic-pricing","tag-claude","tag-claude-sonnet-5","tag-claude-sonnet-5-agent-models-anthropic-pricing-ai-automation-opus-4-8","tag-news","tag-opus-4-8"],"_links":{"self":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/91030","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/comments?post=91030"}],"version-history":[{"count":0,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/posts\/91030\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media\/91031"}],"wp:attachment":[{"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/media?parent=91030"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/categories?post=91030"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.europesays.com\/ai\/wp-json\/wp\/v2\/tags?post=91030"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}