Anthropic's new Sonnet 5 model brings autonomous tool use and competitive pricing while the company expands into scientific research ahead of its IPO.
Anthropic released Claude Sonnet 5 on Tuesday, confirming months of speculation about an upgrade to its mid-tier AI model. The company calls it the most agentic Sonnet yet, capable of planning, using browsers and terminals, and operating autonomously at a level previously reserved for larger, pricier systems. Sonnet 5 is now the default model on Claude's free and Pro tiers and available across Max, Team, and Enterprise plans.
The new model delivers substantial improvements over Sonnet 4.6 across reasoning, coding, and knowledge-work benchmarks, performing close to the flagship Opus 4.8 while costing significantly less to run. Introductory pricing sits at $2 per million input tokens and $10 per million output tokens through August 31, after which standard rates of $3 and $15 take effect. That is less than half the price of Opus 4.8 when accessed through Claude Code, according to Gizmodo.
On safety, Anthropic reports lower rates of hallucination, sycophancy, and other undesirable behaviors than its predecessor, along with improved resistance to prompt-injection attacks. The company noted that Sonnet 5's cybersecurity capabilities remain well below those of its Opus-class and Mythos-class systems, and cyber safeguards are enabled by default as a precaution. Notably absent from the announcement: specific figures on those hallucination improvements, with Anthropic offering only a general claim of lower rates rather than benchmark data Mashable.
The release also made no mention of the model's energy consumption or environmental footprint, a growing concern as models become more capable and computationally intensive. Meanwhile, Anthropic expanded its product lineup with Claude Science, a workbench for researchers that runs on existing models and provides auditable, reproducible results. The offering is part of a broader push to monetize services ahead of a planned initial public offering Yahoo Finance.
Anthropic filed confidential IPO paperwork with the SEC on June 1, followed by OpenAI on June 8. The New York Times has reported OpenAI could delay its debut until 2027. At a San Francisco event Tuesday, Anthropic's life sciences head Eric Kauderer-Abrams said the company will focus on discovering treatments for neglected diseases that traditional biopharmaceutical companies overlook, positioning the effort as a way to build better tools through tight feedback loops with drugmakers CNBC.
The market reaction
Sonnet 5 arrives as developers face mounting pressure to provide cheaper AI tools, driven by the proliferation of AI agents that autonomously handle complex tasks over long time horizons and consume far more tokens than simple chatbots. Both Anthropic and OpenAI have reportedly considered significant price cuts to attract and retain users. Anthropic's strategy of delivering near-flagship performance at mid-tier pricing could reset expectations for what agentic models should cost.
What this means for engineers and buyers
For teams evaluating model providers, Sonnet 5 changes the cost-performance calculus: autonomous tool use and reasoning approaching Opus 4.8 at a fraction of the price makes agentic workflows economically viable at scale. The missing hallucination benchmarks and energy data remain gaps for compliance and sustainability reviews. Anthropic's simultaneous push into scientific workbenches and drug discovery signals a verticalization strategy that could differentiate its platform ahead of public markets.
Anthropic's dual launch of a cheaper agentic model and a science-focused workbench shows a company racing to broaden revenue streams before its IPO while OpenAI weighs a delay. The question now is whether Sonnet 5's pricing holds after August or whether competitive pressure forces another round of cuts.
FAQ
What is the introductory price for Claude Sonnet 5?
Through August 31, Sonnet 5 costs $2 per million input tokens and $10 per million output tokens, then rises to $3 and $15 respectively.
How does Sonnet 5 compare to Opus 4.8?
Anthropic says Sonnet 5 performs close to Opus 4.8 on reasoning, coding, and knowledge-work benchmarks but shows substantially poorer performance on cybersecurity tasks.
What is Claude Science?
Claude Science is a research workbench that lets scientists bring data from disparate sources into one auditable, reproducible environment. It runs on Anthropic's existing models.
When is Anthropic's IPO expected?
Anthropic filed confidential IPO paperwork on June 1 and is on track to go public later this year, though timing remains uncertain.
