OpenAI (ChatGPT)
Text Models
The most stable text models right now are GPT-5.4 and GPT-5.5 — both deliver good results, though they’re not the fastest. If speed matters more to you, go with GPT-5.4 Mini for quicker responses.
One thing to watch out for: models ending in “pro” or “latest” gave me unstructured outputs — random styles and unexpected special characters. I’d avoid those for now.
The models that actually helped me write this article are shown in the screenshot below.
Image Models
- DALL-E 2 — fast, but the images look very “AI-generated”
- GPT-Image-1 or 1.5— better image quality and still reasonably fast; this is the one I recommend
Anthropic (Claude)
The easiest way to pick a Claude model is to just open Claude and see what’s currently stable. As of now, the reliable options are:
- Opus 4.8 — for complex tasks
- Sonnet 4.6 — the sweet spot for everyday use
- Haiku 4.5 — fastest, great for quick answers
Note: Claude doesn’t support image generation yet — maybe in the future! 🙂
Google Gemini
Gemini keeps it simple: all their models are stable for both text and images. Unlike OpenAI, where you can browse old, new, and in-development models all mixed together, Gemini only surfaces stable options in their interface. Less overwhelming, honestly.
OpenRouter
No specific recommendations here — and that’s kind of the point. OpenRouter gives you access to 200+ models, including paid and free versions of OpenAI, Claude, and Gemini models, plus many others.
The free models are worth exploring. Which ones gave you the best results? Drop a comment and let us know!