Models you might see here
No model wins every prompt — this page is a cheat sheet for “try this when…” Names and availability can shift with upstream updates. Want SD-style results without a local GPU? See our Stable Diffusion online guide.
Image generation models
Starting points — your prompt and settings still matter more than the badge on the box
Flux
The reliable default for “I need an image” — strong detail and flexible styles when you describe the scene clearly.
GPT Image Large
When you care about fine texture and crisp edges — still not magic; give it lighting and subject cues.
Seedream
Often good when your prompt is messy but your intent is artistic — worth A/B testing against Flux on the same idea.
Kontext
Useful when your brief has several constraints at once — it tries harder to keep them all in frame.
Text generation models
Different strengths: long answers, careful tone, tooling, or raw code
GPT-5
Generalist for drafting, explaining, and refactoring — sanity-check anything important.
Claude
Often strong for careful prose, long context, and “explain this like I’m tired” moments.
Gemini
Gemini-flavored stacks may bring search, URL fetch, or code execution when the route supports it — check the live UI.
DeepSeek V3.2
A solid chat/workhorse — try it when you want fast back-and-forth without overthinking the brand name.
Qwen3-Coder-30B
Bias toward code: scaffolding, refactors, and “why does this break?” questions — always run tests yourself.
Video & audio
Latency and limits hit hardest here — read upstream notes before you promise a client a date
Seedance
Text-to-video path in the ecosystem — expect iteration cost in time, not just tokens.
Veo (Alpha)
High aspirations, alpha reality — treat outputs as experiments until the upstream says otherwise.
AlphaOpenAI Audio
Voice in and out — pick the voice that fits your clip, then normalize levels like a human.
Pick one and stress-test it
Use the web tools if you just want a feel — open the API page when you are wiring production