Serverless specialty GB

Stability AI Platform.

Stability AI's first-party API for Stable Diffusion + audio/video models.

At a glance

Service type: Serverless specialty
Trust tier: Tier 2
Headquarters: GB
OpenAI-compat: No
Open weights: Yes
Proprietary: Yes

When to pick Stability AI Platform

Best for

Ultra-low-latency inference (Groq's LPU silicon, Cerebras).
Image / video / audio generation via per-second billing.
Workloads where the specialty's hardware advantage outweighs cost.

Avoid for

General LLM workloads where a generalist aggregator is cheaper.
Workloads needing feature parity across many models.

Models on Stability AI Platform

Pricing + measured speed + self-host alternative, one row per model. Click a column header to sort.

1 models · 0 benchmarked

Model ↕	Maker ↕	Access ↕	$/M in ↕	$/M out ↕	Tokens/sec ↕	TTFT ↕	Self-host on ↕
Stable Diffusion 3.5 Large	Stability AI	api direct	—	—	—	—	1× Nvidia GeForce RTX 2060 · INT4	Open →

Peers in the same bucket

Sanity-check before you commit.

Groq 5

fal.ai 4

Black Forest Labs API 1

Ollama 44

FAQ

Frequently asked.

How does Stability AI Platform bill for inference?

Most inference providers bill per million input + output tokens. Some (like Replicate, fal.ai) use per-second billing where you pay for actual compute time on the underlying GPU. Per-token rates are listed on each model's access row.

Does Stability AI Platform offer an OpenAI-compatible API?

Not by default. Use Stability AI Platform's native SDK — most clients support a thin OpenAI compatibility wrapper if you need it.

Which models does Stability AI Platform host?

See the model list on this page. Each row links to a per-model show page where you can compare across every provider that carries the same SKU.

What's the difference between Stability AI Platform and a GPU rental host?

Stability AI Platform hosts inference for you — you pay per token, the provider runs the GPU. GPU rental (Vast.ai, Lambda, RunPod) gives you the raw GPU at $/hr and you run the model yourself. Both have catalog pages on RentGPU; pick based on whether you want managed inference or full control.