Anthropic launches Claude Sonnet 4.5, its best AI model for coding

On Monday, Anthropic launched a caller frontier exemplary called Claude Sonnet 4.5, which it claims volition state-of-the-art connected coding benchmarks. The institution says Claude Sonnet 4.5 is susceptible of gathering “production-ready” applications, a leap successful reliability from erstwhile AI models.

Claude Sonnet 4.5 volition beryllium disposable via the Claude API and successful the Claude.ai chatbot. The pricing for developers is the aforesaid arsenic Claude Sonnet 4: $3 per cardinal input tokens (roughly 750,000 words, oregon much than the full Lord of The Rings series) and $15 per cardinal output tokens.

In the past year, Anthropic’s AI models person emerged arsenic a favorite among developers and enterprises, successful ample portion owed to their beardown show connected bundle engineering tasks. Apple and Meta reportedly usage Claude AI models internally, and Anthropic has made a important concern selling API entree to AI coding applications specified arsenic Cursor, Windsurf, and Replit. Recently, OpenAI’s GPT-5 has challenged Anthropic’s dominance successful the space, outperforming Claude models connected a assortment of coding benchmarks.

Anthropic says Claude Sonnet 4.5 offers industry-leading show connected respective coding benchmarks, including SWE-Bench Verified. However, Anthropic AI researcher David Hershey tells TechCrunch that it is hard to seizure Claude Sonnet 4.5’s show connected benchmarks alone.

Hershey says he’s seen Claude Sonnet 4.5 codification autonomously for up to 30 hours during aboriginal trials with immoderate endeavor customers. In that time, helium watched the AI exemplary not lone physique an application, but basal up database services, acquisition domain names, and execute a SOC 2 audit to marque definite the merchandise was secure.

In a connection shared with TechCrunch Cursor CEO Micheal Truell said Claude Sonnet 4.5 represents state-of-the-art coding performance, specifically connected longer skyline tasks. Windsurf CEO Jeff Wang said successful a connection that Claude Sonnet 4.5 represents a “new procreation of coding models.”

Anthropic besides claims that Claude Sonnet 4.5 is its astir aligned frontier AI exemplary yet, with little rates of sycophancy and deception than erstwhile models. The institution says it has besides improved Claude’s susceptibility to punctual injection attacks.

Alongside the motorboat of Claude Sonnet 4.5, Anthropic is besides launching the Claude Agent SDK. The institution says this is the aforesaid infrastructure that powers Claude Code, and tin beryllium utilized to assistance developers physique their ain agents.

Anthropic is besides releasing a impermanent probe preview called “Imagine with Claude” for Max subscribers, which shows the AI exemplary generating bundle connected the fly. The institution says the exemplary volition respond to idiosyncratic requests successful existent time, with nary predetermined functionality oregon prewritten code.