An up to date comparison of the best AI models
Last updated October 5, 2026.
MODELS
The current top model is Claude Fable 5.1, but Claude Opus 5.5 is so close to it that there is no reason to pay extra to use Fable. The new Sonnet 5.5 is a great workhorse that I use for everything practical. Right behind Claude is OpenAI’s GPT-6 family, with Astra as the flagship and 6.1 Sol as the workhorse. Google is now officially behind, but we are expecting Gemini 4 any day now.
There are other models (Grok 4.7, GLM 5.3, Kimi K3), but you don’t need them.
GENERAL TOOLS
AI is more than just the models. It is also the tools or scaffolding they come with. Claude has agentic tools like Cowork and Code, and integrations in for example Chrome, Excel, PowerPoint, Outlook etc. ChatGPT now also has such integrations, and recently added “Work” to the existing “Codex” to match Claude’s Cowork and Code. ChatGPT Work so far appears to be on par with Claude Computer and Perplexity Computer.
In a complete reversal from earlier in the year Claude is now very generous with the usage you get for your $20/month (I use Opus 5.5 all day and barely run out by the end of week) while ChatGPT is not (Astra is greedy, Sol a little less so). Gemini is integrated in all the Google products (as long as you pay the $20/month), but really the one tool that matters is Gemini Notebook (just renamed from NotebookLM).
Then there is of course the new battle of the assistants: Meta’s Muse Assistant made a big splash and is probably what most people will try first (it is free). Grok Bot gets great reviews, but most people don’t use Grok. OpenAI launched Dots yesterday and it looks good.
SPECIALIZED TOOLS
The image generation crown changed hands in April: ChatGPT’s GPT Image 2.5 now tops the blind-vote leaderboards, especially for readable text, diagrams, and layouts. For photorealism and high-res editing (up to true 4K), Google’s Nano Banana is still up there, but I find myself using GPT Image for everything except when I work in Notebook.
I don’t make much video but Google Omni seems very good based on what little I’ve tried with it. See for example this from the prompt “me riding a unicorn through the swamp”:
If you are looking for a resource for learning how AI works and how to work with it, I recommend my Tulane Applied AI Literacy course that is available for free on YouTube.
New to AI and not sure where to start? What Can You Actually Use AI For? sorts the tasks AI is good at into five categories, with everyday examples.
