Image: Anthropic
Anthropic may be using its specialization in coding and STEM as a differentiator in the buzzy frontier model market.
Anthropic has thrown down another gauntlet in the AI fight with Claude Sonnet 4.5, a model that the company claims excels at coding. Claude Sonnet 4.5 is claimed to perform better than its predecessors at writing production-ready code, using computers, performing cybersecurity tasks, and researching, modeling, and forecasting for financial services.
“Claude Sonnet 4.5 resets our expectations — it handles 30+ hours of autonomous coding, freeing our engineers to tackle months of complex architectural work in dramatically less time while maintaining coherence across massive codebases,” said Sean Ward, chief executive officer and co-founder of iGent AI, in Anthropic’s announcement.
Enabling an AI to work independently for so long suggests Anthropic might deliver on the promise of an AI coworker. Claude Sonnet 4.5 outperforms both Claude Opus 4.1 — Anthropic’s previous most powerful model, released only two months ago — on a variety of benchmarks, the company said. Alongside sustained coding, Claude Sonnet 4.5 also is claimed to excel at tasks in law, medicine, and broader STEM fields.
Claude Sonnet 4.5 was tuned to avoid sycophancy, scheming, power-seeking, and encouraging delusional thinking. Anthropic calls it its “most aligned” model.
It is priced at $3 per million input tokens and $15 per million output tokens, the same as Claude Sonnet 4 — a good deal for existing users who require heavy-duty coding, though relatively pricey in the broader generative AI landscape.
Anthropic also released upgrades across its products:
Gartner forecasts 149.8% year-over-year growth in generative AI spending in 2025, reaching about $14.2 billion in end-user spending.
As the models improve, the differences between them become less dramatic. Claude Sonnet 4.5 shows a slight improvement over its predecessors, according to a research note from Gartner (link for subscribers only).
“With the launch of Claude Sonnet 4.5, Anthropic is seeking to reinforce its leadership in the coding domain,” said Gartner analyst Arun Chandrasekaran in an email to TechRepublic. “However, in addition to the coding domain they have also trained the model to perform competently in two other areas where they see opportunities, finance and cybersecurity. However, the improvements across these domains are incremental.
“However, on the back of its success in establishing MCP as an industry protocol for Gen AI and Agentic AI communication, Anthropic has added significant capabilities in memory, agent coordination and autonomy.”
Anthropic’s new model differentiates itself through its specializations, particularly its focus on coding and finance, two industries that may undergo significant job transformations due to generative AI.
“Claude Sonnet 4.5 signals Anthropic’s intention to shift toward domain specialization in an increasingly competitive generative AI (GenAI) model market landscape,” the research note said.
In particular, the cybersecurity skills of Claude Sonnet 4.5 validate strong market demand for LLMs that can perform threat analysis and security testing, Gartner said.
The White House’s latest plan for reducing U.S. dependence on foreign technology suppliers calls for a 1:1 ratio of imported to domestic chips.
Megan Crouse has a decade of experience in business-to-business news and feature writing, including as first a writer and then the editor of Manufacturing.net. Her news and feature stories have appeared in Military & Aerospace Electronics, Fierce Wireless, TechRepublic, and eWeek. She copyedited cybersecurity news and features at Security Intelligence. She holds a degree in English Literature and minored in Creative Writing at Fairleigh Dickinson University.
