Quotaflow Model Library — supported AI models by use case
Choose the right Quotaflow model for chat, agents, code, embeddings, image generation, image editing, and video generation. The catalog below is organized like a product model library: start with your use case, pick a curated model family, then open the endpoint-specific reference.
What is the Quotaflow Model Library?
The Model Library is the customer-facing list of model ids and endpoint families that Quotaflow documents for production integrations. It explains what each model is good for, which protocol surface it uses, and which reference page has copy-pasteable API examples.
Your authenticated GET /models response is still the source of truth for your key. A package, beta flag, or product-scoped key can make your returned list smaller than the public catalog.
Why use Quotaflow model routing?
Quotaflow lets teams keep official-style request shapes while centralizing API keys, team usage, package limits, and model access in one control plane. That means you can point OpenAI-compatible, Anthropic-compatible, Gemini-compatible, and creative-media clients at Quotaflow without teaching every app a new billing or credential system.
Use this page when you need to answer three practical questions before shipping:
- Which model id should I configure in my SDK?
- Which endpoint should my client call?
- Which models are best for my latency, quality, and cost constraints?
Curated models
| Category | Start with | Endpoint family | Best for | Reference |
|---|---|---|---|---|
| GPT agents | gpt-5.4 | /openai/v1/responses | Production agents, tool use, app reasoning | Responses API |
| Fast GPT chat | gpt-5.4-mini | /responses, /chat/completions | UI chat, extraction, summaries, routing | OpenAI-compatible models |
| Lightweight GPT routing | gpt-5.4-nano | /responses, /chat/completions | Cost-sensitive simple chat, extraction, and routing | OpenAI-compatible models |
| Coding agents | gpt-5.3-codex | /responses | Codex-style repository work and agent tasks | Connect Coding Agents |
| Embeddings | text-embedding-3-small, text-embedding-3-large, text-embedding-ada-002 | /embeddings | Search, RAG, clustering, semantic matching | Embeddings |
| OpenAI-style images | gpt-image-2 | /images/generations, /images/edits | Product images, masked edits, marketing creative | GPT Image 2 |
| Fast image iteration | gemini-3.1-flash-image-preview | /images/*, Gemini :generateContent | Drafts, reference-image edits, many variants | Gemini 3.1 Flash Image |
| Higher-quality image work | gemini-3-pro-image-preview | /images/*, Gemini :generateContent | Final creative candidates and detailed prompts | Gemini 3 Pro Image |
| Veo video generation | veo-3.1-fast-generate-001 | /videos | Veo-compatible async text-to-video and image-to-video tasks | Veo 3.1 Fast |
| Seedance video generation | seedance-2.0-fast | /videos | Seedance 2.0 text, image, portrait, and reference-media video tasks | Seedance 2.0 Fast |
| Claude-compatible clients | claude-sonnet-5, claude-sonnet-4-6, claude-opus-4-6, claude-sonnet-4-5, extended ids returned by /models | /v1/messages | Claude-style coding and reasoning clients | Messages |
| GLM chat | glm-5.2 | /openai/v1/chat/completions | OpenAI-compatible chat clients that need GLM text output; supports messages, streaming, and tools; strict JSON mode returns 400 invalid_request_error | Chat Completions |
| Kimi coding chat | kimi-k2.7-code | /openai/v1/chat/completions | Kimi text for coding, agent, long-context chat, streaming, tools, and strict JSON mode | Chat Completions |
| Gemini-compatible clients | gemini-3.5-flash, gemini-3.1-pro-preview, gemini-2.5-pro | /v1beta/models/{model}:generateContent, /openai/v1/chat/completions | Gemini text, multimodal, and reasoning-capable apps; Gemini-native features stay on generateContent/countTokens | generateContent |
Key features
- Protocol-compatible endpoints: keep OpenAI-compatible, Anthropic-compatible, or Gemini-compatible client shapes where possible.
- Key-scoped discovery:
GET /modelsreturns only the model ids enabled for the calling key. - Creative media coverage: use the same Quotaflow account for text, embeddings, images, and video tasks.
- Model-specific references: image and video pages include endpoint examples, parameter guidance, failure modes, and production checklists.
- Fast unsupported-model errors: invalid or unavailable model ids fail clearly instead of being silently substituted.
How to choose a model
- Pick the protocol your client already speaks: OpenAI-compatible, Anthropic-compatible, or Gemini-compatible.
- Call the matching
/modelsendpoint with the exact key your app will use. - Choose the first curated model in the table for your category.
- Send a small non-streaming test request.
- Add a streaming, multipart, or async polling test if your production flow uses that capability.
Pricing and access
Model availability depends on your Quotaflow package and enabled product surfaces. If a model appears in this catalog but not in your authenticated /models response, ask your Quotaflow admin to enable that model family for the key or create a key scoped to the right product.
For creative media workloads, budget for longer runtimes than text calls. Image and video jobs should use client timeouts, retry policies, and UI pending states that match media-generation latency.
Use cases
| Use case | Recommended path |
|---|---|
| Build a coding-agent gateway | gpt-5.3-codex with /responses, then follow Codex setup. |
| Add image editing to an ecommerce tool | gpt-image-2 with /images/edits and masks for localized changes. |
| Create many visual concepts quickly | gemini-3.1-flash-image-preview for fast variants, then optionally promote final candidates to a higher-quality image model. |
| Generate short social clips | veo-3.1-fast-generate-001 or seedance-2.0-fast with async /videos tasks and a polling UI. |
| Keep an existing Claude client | Use the Anthropic-compatible /v1/messages surface and a Claude-compatible model returned by your key. |
| Add GLM text chat | Use glm-5.2 through /openai/v1/chat/completions after confirming it appears in /models; use messages, streaming, and tools, but do not send strict JSON mode. |
| Add Kimi coding chat | Use kimi-k2.7-code through /openai/v1/chat/completions after confirming it appears in /models; streaming, tools, and strict JSON mode are supported on this chat path. |
FAQ
How do I know which models my key can use?
Call GET https://api.quotaflow.ai/openai/v1/models for OpenAI-compatible keys, or the protocol-specific discovery endpoint for Anthropic-compatible and Gemini-compatible clients. Use the same key and base URL that your production app will use.
Are image and video models OpenAI-compatible?
Quotaflow exposes OpenAI-compatible image endpoints for image generation and edits. Video generation uses an OpenAI-style task endpoint where the create call returns a task id and the client polls for completion.
Do all keys include every model?
No. Keys can be scoped by product, package, customer, or beta access. Treat this page as the public catalog and /models as the exact enabled list for a specific key.
What happens if I send an unsupported model id?
The request fails with a model or configuration error. Update the configured model id or enable the model for the key; do not retry unsupported-model errors as transient failures.
Next steps
- Configure the OpenAI-compatible base URL.
- Build an image flow with GPT Image 2.
- Build a video flow with Veo 3.1 Fast or Seedance 2.0 Fast.
- Give coding agents the Codex setup guide.