Some models that I’ve been using recently that I find surprisingly good:

- DeepSeek for planning code tasks https://www.deepseek.com
- Phind for visual answers and multi-step reasoning https://www.phind.com
- Perplexity Deep Research for searching resources and in-depth research https://www.perplexity.ai

🇨🇦 • Explorer. Code, design, and mountains. Building Fern https://fernhq.com Previously Shopify

I too replaced Claude with DeepSeek for this planning step. Will have to try more with o3

Building Higher apparel at High Line 🧢 
Host at /clickcast /cpg || Husband to @shiwen

I did notice deepseek didn’t do better with code, but I did see that OpenAI’s recent models do quite well with code. Just haven’t had the time to test it recently

I usually try both (deepseek and o3) for comparison and get consistent better results with deepseek. For coding act step sonnet is the best