• Daily AI Skills
  • Posts
  • Claude 3.7 Sonnet: Is it the best Coding LLM out there?

Claude 3.7 Sonnet: Is it the best Coding LLM out there?

PLUS: Perplexity Comet & Qwen's New Thinking Model

Author

Daily AI Skills
February 25, 2025

Welcome back to Daily AI Skills.

Here’s what we are covering today:
1. Claude 3.7 Sonnet - The Best AI Coding Assistant
2. Perplexity Comet - Agentic Browser Use
3. Qwen QwQ Max - New Open-Source Thinking Model

Claude 3.7 Sonnet - Anthropic’s First Hybrid Reasoning Model

Anthropic has unveiled Claude 3.7 Sonnet, a groundbreaking AI model that introduces hybrid reasoning—blending rapid responses with in-depth, controllable thinking.

Source: Anthropic

Key Highlights about the model:

  • Claude 3.7 Sonnet allows users to switch between standard and "extended thinking" modes, enabling deeper self-reflection on complex queries.

  • API users can adjust thinking budgets up to 128K tokens, balancing speed, cost, and response quality.

  • The model achieves state-of-the-art (SOTA) performance in instruction-following, reasoning, and multimodal tasks, outperforming competitors like o1, o3-mini, and DeepSeek R1.

  • Extensive evaluations show improvements in handling complex codebases, solving real-world software issues, and maintaining superior accuracy in structured problem-solving.

  • Pricing remains unchanged from its predecessor, keeping it accessible while enhancing capabilities.

Perplexity Comet - Agentic Browser Use

Perplexity is developing a new web browser called Comet, aiming to redefine how users interact with AI-powered search. The company has launched a sign-up page for early access, but details on its release date and features remain scarce.

Key Highlights of Comet:

  • Perplexity describes Comet as a browser built for “agentic search,” suggesting deep AI integration for more dynamic web navigation.

  • The browser enters a competitive market dominated by Chrome and emerging AI-powered alternatives like OpenAI Operator

  • Perplexity has been rapidly expanding, launching a deep research tool this month and an AI search API in January.

Qwen’s New Open-Source Thinking Model

Alibaba's Qwen team has unveiled QwQ-Max-Preview, an advanced AI model focused on deep reasoning, enhancing capabilities in problem-solving and agentic tasks—while also hinting at an upcoming open-source release.

Key Highlights of the Model:

  • Built on Qwen2.5-Max, the model significantly improves reasoning abilities, excelling in mathematics, coding, and complex workflows.

  • "Thinking (QwQ)" mode allows users to view the AI’s thought process in real time, offering transparency in decision-making.

  • Open-source plans: Qwen has announced that both QwQ-Max and Qwen2.5-Max will soon be available under an Apache 2.0 license, allowing developers full access.

  • Lightweight variants like QwQ-32B are also in the works, enabling local deployment on devices with limited computing power.

AI just levelled up—whether it’s thinking deeper, searching smarter, or going open-source, the future is looking a lot less artificial and a whole lot more intelligent.

Share it with people you know are keeping pace with the changing AI world and stay tuned for the next edition to stay ahead of the curve!