Reference · June 2026
For every provider: where to find it, where to sign up, and which requested models it really supports. The top split is purely about onboarding friction — every platform here gives free credit, but the first group makes you put a credit card on file to claim it, and the second doesn't. Video & audio tools sit in their own group.
GPT-5.5 & GPT image 2 live on OpenAI, Azure, Vercel — and, text-only, AWS Bedrock. Never on Google Cloud.
Nano Banana 2 & Veo live on Google AI + Vertex AI only. (Veo is also resold inside Runway.)
Opus distributes very widely; DeepSeek V4 Pro & Kimi K2.6 are open-weight and turn up on most open-model hosts.
All of these give free credit — but you have to put a card on file to claim it. Sorted into tiers below.
Hyperscalers and major clouds.
Smaller clouds and inference specialists.
relaxAI is Civo's own OpenAI-compatible service running on Civo-owned GPUs — usage sits inside the Civo ecosystem (API via the relaxAI Plus plan). Raw GPU/K8s self-hosting is also available.
Card-gated credit, but no compelling first-party model API — listed for completeness only.
No model-support links — there's no relevant model catalog to point at.
Same free credit, but you can start without entering any card details.
Strong model access with no card required to start.
Inference specialists, no card to start — Fireworks adds the AMD $50 route.
Join the AMD AI Developer Program and choose the Fireworks AI credit option (partner page).
Little or no usable model API — listed for completeness only.
No model-support links — there's no relevant model catalog to point at.
Generative media platforms — grouped separately from the credit-card split.
Generative video / audio platforms and the models they expose.