Anthropic debuts Claude Sonnet 4.5, its top-tier AI model for developers.

On Monday, Anthropic introduced Claude Sonnet 4.5, a new frontier model it claims delivers top-tier performance on coding benchmarks. The company states this model can create "production-ready" applications, not just prototypes, marking a significant leap in reliability over previous AI systems.

Claude Sonnet 4.5 will be accessible through the Claude API and within the Claude chatbot. Developer pricing remains consistent with Claude Sonnet 4: $3 per million input tokens (approximately 750,000 words) and $15 per million output tokens.

Over the past year, Anthropic's AI models have become a preferred choice for developers and enterprises, largely due to their robust performance on software engineering tasks. Companies like Apple and Meta are reported to use Claude models internally, and Anthropic has built a substantial business providing API access to AI coding platforms such as Cursor, Windsurf, and Replit. Recently, OpenAI's GPT-5 has emerged as a strong competitor, outperforming Claude models on various coding benchmarks.

Anthropic asserts that Claude Sonnet 4.5 achieves industry-leading results on several coding benchmarks, including SWE-Bench Verified. However, Anthropic AI researcher David Hershey notes to TechCrunch that benchmark scores alone don't fully capture the model's capabilities.

Hershey reports that during early trials with enterprise clients, he observed Claude Sonnet 4.5 coding autonomously for up to 30 hours. In these sessions, the model not only built applications but also set up database services, purchased domain names, and conducted SOC 2 audits to ensure product security.

In a statement to TechCrunch, Cursor CEO Micheal Truell described Claude Sonnet 4.5 as representing the cutting edge in coding performance, particularly for extended, complex tasks. Windsurf CEO Jeff Wang stated the model signifies a "new generation of coding models."

Anthropic also claims Claude Sonnet 4.5 is its most aligned frontier AI model to date, exhibiting lower rates of sycophancy and deception than its predecessors. The company adds it has enhanced the model's resistance to prompt injection attacks.

Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just a few of the 250+ industry leaders hosting 200+ sessions packed with insights to drive startup growth and sharpen your competitive edge. Don't miss the 20th anniversary of TechCrunch and this opportunity to learn from the top minds in tech. Secure your ticket early to save up to $444.

Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025

Netflix, Box, a16z, ElevenLabs, Wayve, Hugging Face, Elad Gil, Vinod Khosla — just a sample of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don't miss a chance to learn from the top voices in tech. Grab your ticket before doors open to save up to $444.

Alongside Claude Sonnet 4.5, Anthropic is launching the Claude Agent SDK. The company says this infrastructure, which powers Claude Code, will help developers build their own AI agents.

Anthropic is also releasing a temporary research preview, "Imagine with Claude," for Max subscribers. This feature demonstrates the AI model generating software in real-time, responding to user requests without relying on prewritten code or predetermined functions.

The intense competition in AI has led companies to release flagship models every few months. Claude Sonnet 4.5 arrives less than two months after Anthropic's previous model, Claude Opus 4.1. These rapid release cycles make it challenging for any single company to maintain a decisive lead for long.