The race to build intelligent software assistants just hit a new milestone as automated capabilities become standard across the technology sector. To lead this shift, Anthropic just launched Claude Sonnet 5, a cheaper yet highly capable upgrade to its mid-tier software intelligence. The lab designed this specific version to operate autonomously, executing complex plans without constant human direction.
According to an official company announcement, Sonnet 5 can independently create multi-step schedules, interact directly with developer interfaces, and execute code within isolated environments. Just a few months ago, achieving this level of freedom required much bulkier and more expensive language frameworks. This launch mirrors recent product releases from major competitors like OpenAI and Google. OpenAI previewed its highly automated GPT-5.5 Sol system last week, which splits massive projects among smaller digital sub-agents to complete long-term goals without supervision. Similarly, Google built its Gemini 3.5 Flash upgrade to act as an automated builder that creates, tests, and adjusts real-world code while keeping human interaction to a bare minimum.
This aggressive rollout pattern proves that independent task execution is no longer a premium feature. It is now the baseline standard across every single pricing bracket in tech. The core competition is shifting from what these models can do to how cheaply and reliably they can execute those tasks without breaking down.
Sonnet 5 targets the high performance levels of Anthropic’s premium Opus 4.8 framework but charges a tiny fraction of the price. Starting immediately, Sonnet 5 operates as the default engine for all free and premium subscribers on the platform. To attract high-volume app developers, Anthropic set an introductory price of two dollars per million input tokens and ten dollars per million output tokens through August 31. After that promotional window closes, output prices will rise slightly to fifteen dollars per million tokens. These rates make Sonnet 5 significantly more affordable than Opus 4.8, though it still costs more than Google’s budget-friendly Gemini 3.5 Flash engine.
The fresh release delivers substantial leaps over its predecessor, Sonnet 4.6, which debuted back in February. Independent performance tests show massive growth in core areas like reasoning, complex programming, and general research. For example, during specialized evaluation benchmarks, Sonnet 5 scored an impressive 63.2% on independent coding assignments, closely chasing the 69.2% record held by the premium Opus 4.8 and easily beating the 58.1% score of the older Sonnet 4.6. On comprehensive knowledge tasks, Sonnet 5 actually beat the expensive Opus 4.8 model, proving highly effective at solving abstract challenges that require subtle judgment calls.
This mix of low operational costs and high intelligence makes it a top option for everyday workflows. Daniel Shepard, a senior developer at automation platform Zapier, shared that his team handed Sonnet 5 a difficult multi-stage assignment involving updating data inside customer management software while simultaneously blasting out email alerts. The model successfully completed the entire job from start to finish without stalling midway, a task that older software configurations regularly failed to execute.
The security architecture received a major overhaul as well. The new framework exhibits fewer undesirable tendencies, refusing to cooperate with deceptive human prompts or malicious tricks. Testing data proves it excels at blocking prompt-injection hacks and rejects unsafe requests much more consistently than previous versions, giving enterprise clients a reliable tool for automated production lines.