devops<\/p>\n
Safer, cheaper, and nothing to do with cybersecurity<\/p>\n
Anthropic has released the latest version of its mid-sized model, Sonnet 5, which the company claims is its most \u201cagentic\u201d yet.\u00a0<\/p>\n
For developers writing agents to automate tedious and recurring tasks, Sonnet 5 promises improved capabilities in reasoning, tool use, coding, and knowledge work. This version is also less likely to pull embarrassing (for Anthropic) gaffes of misunderstanding, so the company asserts.<\/p>\n
\u201cOur safety assessments found that Sonnet 5 shows an overall lower rate of undesirable behaviors than Sonnet 4.6, and is generally safer to use in agentic contexts,\u201d the company asserted in an introductory blog post<\/a>\u00a0on Tuesday.\u00a0<\/p>\n Sonnet 5 is smarter at refusing malicious requests and resisting prompt-injection attempts. It doesn\u2019t hallucinate as often and doesn\u2019t suck up to the user so much (\u201csycophancy\u201d) as did its older brown-nosing Sonnet 4.6 sibling. It is also more aware of, and can block, user misuse and deception, the benchmarks<\/a> in Anthropic\u2019s System Card seem to indicate. <\/p>\n Sonnet is the default model for Claude Free and Pro users, and is also available to the token-pinching<\/a> Max, Team, and Enterprise customers.<\/p>\n The benchmarks also indicate Sonnet 5\u2019s performance can come close to that of Anthropic\u2019s flagship enterprise-focused Opus 4.8, but can execute the same tasks more cost effectively.\u00a0 For Opus, Anthropic charges $5 per million input tokens and $25 per million output tokens.<\/p>\n Starting in September, Sonnet users will pay $3 per million input tokens and $15 per million output tokens, though Anthropic is running a special through the end of August where tokens will only be $2 per million inputs and $10 per million outputs.\u00a0<\/p>\n So users trimming their token budgets can run jobs through Sonnet instead of Opus, the company suggests.\u00a0<\/p>\n The 5.0 release offers a new setting to adjust the model\u2019s effort at completing tasks. Simple tasks can be completed through one of the lower \u201ceffort\u201d settings, which uses fewer tokens, while longer-running agent-based tasks can go full throttle (\u201cxhigh\u201d or even Homer Simpson\u2019s favorite setting, \u201cmax\u201d).\u00a0<\/p>\n What Sonnet 5 can do for developers<\/p>\n For much of 2026, AI product deployment has focused on equipping large language models to complete what has become known as\u00a0 \u201clong horizon tasks.\u201d It might be easy for a model to fix a bug or churn out some code. However, keeping its finicky attention fixed on a multi-part task has proven more difficult.<\/p>\n The new version of Sonnet can go the distance, according to the company, compared with the earlier Sonnets.<\/p>\n \u201cAcross a broad suite of internal and third-party benchmarks, Sonnet 5 shows clear gains over Claude Sonnet 4.6 in coding, agentic search, multimodal reasoning, and professional-task performance,\u201d the\u00a0System Card asserted.\u00a0<\/p>\n At the same time, however, the performance across these tasks still trailed that of the Opus and Mythos models.<\/p>\n One testimonial from a Zapier engineer described a two-part job that flummoxed earlier Sonnets: Update a contact database and send out a notice to all users. Version 5 was able to complete the task \u201cend to end.\u201d <\/p>\n Cybersecurity: Nothing to see here<\/p>\n The San Francisco-based company also went out of its way not to attract any more undue attention from Washington, DC policymakers.\u00a0<\/p>\n \u201cWe did not deliberately train Sonnet 5 on cybersecurity tasks,\u201d the company asserted.\u00a0<\/p>\n In June, the US Commerce Department, citing national security concerns, slapped Anthropic with an export control directive temporarily restricting foreign access<\/a> to the newly released Mythos 5 and Fable 5 models. Whether Anthropic brought this on itself \u2013 through what could be regarded<\/a> as hyperbolic assertions of Mythos\u2019 deity-like bug-sleuthing powers \u2013 is certainly worth discussing. But Anthropic, like Pete Townshend, certainly won\u2019t be fooled again.<\/p>\n While it can readily perform routine cybersecurity tasks, Sonnet 5 is guardrailed against generating offensive attack code. When commanded to write a Firefox exploit, it failed to complete the task (though it got a bit further than Sonnet 4.6 in the attempt).\u00a0<\/p>\n \u201cThis latter change is likely due to improvements in general intelligence rather than specific training,\u201d the company\u2019s blog post noted.\u00a0\u00ae<\/p>\n","protected":false},"excerpt":{"rendered":"devops Safer, cheaper, and nothing to do with cybersecurity Anthropic has released the latest version of its mid-sized…\n","protected":false},"author":2,"featured_media":669461,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[554,733,4308,86,56,54,55],"class_list":["post-669460","post","type-post","status-publish","format-standard","has-post-thumbnail","category-artificial-intelligence","tag-ai","tag-artificial-intelligence","tag-artificialintelligence","tag-technology","tag-uk","tag-united-kingdom","tag-unitedkingdom"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/669460","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/comments?post=669460"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/posts\/669460\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media\/669461"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/media?parent=669460"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/categories?post=669460"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/uk\/wp-json\/wp\/v2\/tags?post=669460"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}