Clik here to view.
Anthropic has announced the release of Claude 3.7 Sonnet, its latest AI model, described as the industry's first 'hybrid AI reasoning model.' Unlike traditional AI chatbots that require users to pick between speed and accuracy, Claude 3.7 Sonnet offers both real-time responses and more deliberate, reasoned answers. Users can choose to activate the model's 'reasoning' mode, prompting it to take more time to process complex queries.
Image may be NSFW.
Clik here to view./socialsamosa/media/post_attachments/67978ecb-1b1.png)
The rollout begins on Monday, with the model available to all users and developers. However, the reasoning features will be limited to those subscribed to Anthropic's premium Claude chatbot plans. Free users will have access to the standard, non-reasoning version. According to Anthropic, Claude 3.7 Sonnet surpasses its predecessor, Claude 3.5 Sonnet, in performance.
Pricing is set at $3 per million input tokens and $15 per million output tokens. While more expensive than OpenAI's o3-mini and DeepSeek's R1, both strictly reasoning models, Anthropic highlights that Claude 3.7 Sonnet integrates multiple capabilities into a single model.
Claude 3.7 Sonnet features a 'visible scratch pad' to reveal its internal thought process for most prompts, with some portions redacted for trust and safety. The model scored 62.3% on the SWE-Bench test for coding tasks, outperforming OpenAI's o3-mini at 49.3%. It also achieved 81.2% on the TAU-Bench test for AI interaction with simulated users, compared to OpenAI's o1 model at 73.5%.
Image may be NSFW.
Clik here to view./socialsamosa/media/media_files/2025/02/25/qm59HCA8uaesyrzu64uh.webp)
Alongside the new AI model, the company is launching Claude Code, an agentic coding tool available as a research preview. Developers can use plain English commands to modify codebases, test for errors, and push projects to GitHub. Access to Claude Code will initially be on a 'first come, first serve' basis.
Anthropic's latest releases come amid increasing competition in AI development. The company aims to integrate reasoning seamlessly into AI models, with plans for future versions of Claude to determine how long to 'think' without user input.
