Which reasoning model to use and for what to get most of your time and money?

After the GPT-4.5 release, it seems CoT reasoning models are the only way forward for improved LLM performance. (Are the base models hitting the wall?) And we now have 4-5 labs with frontier thinking/reasoning models: OpenAI with o3-mini-high, Anthropic with Claude 3.7 Sonnet thinking, XAI with Grok 3(think), and Deepseek with r1. (I didn't add Gemini because it's not good enough) There are too many options, and though it’s a nice place to be, choosing which ones to get the most of your time and money might still be confusing. (I don't particularly like juggling between models) So, I did a little test of all these frontier reasoning models on a personal set of questions across complex reasoning, math, coding, and writing to determine which one to go for which tasks. Here’s the overall findings * **Grok 3 Think** is the top choice for coding and mathematical validation. Also, the biggest plus is Grok is honest and does what he is asked. Least judgemental. * **Claude 3.7 Sonnet** is the best **all-rounder**, especially for writing and reasoning. You cannot go wrong with either Sonnet or Grok 3 for coding. * **DeepSeek R1** is ideal for **math-heavy tasks** but struggles elsewhere. * **OpenAI o3-mini-high** is **fast & efficient** but lacks fine-tuned reasoning in some areas. For a full breakdown of my analysis, check out this blog post: [**Comparative Analysis of CoT Reasoning Models**](https://composio.dev/blog/cot-reasoning-models-which-one-reigns-supreme-in-2025/) I personally find myself using Claude 3.7 Sonnet more than anything. It writes good code (though it feels overfit for coding sometimes), writes decent prose, and reasons concisely. I would love to know the reasoning models you find most beneficial for your tasks.