Modelflare Blog

Introducing Modelflare: one API key for text, image, and video models

Modelflare is an API gateway that serves text, image, and video models behind one key, with OpenAI-compatible, Anthropic, and Gemini endpoints and per-request cost records.

Diagram: one entry point routed to several model backends

Modelflare is a unified API gateway for text, image, and video models. One API key covers OpenAI-compatible chat, the Responses API, Anthropic Messages, image generation, and asynchronous video. Usage and cost are recorded for every request. This post covers the endpoints, how keys and routing groups work, and how to send a first request.

What the gateway provides

Chat Completions, Responses, Anthropic Messages, Images, and Video share the same authentication and routing groups. Applications built on the OpenAI SDK change the base URL to https://modelflare.dev/v1 and keep the rest of their code. Image and video families that have their own request formats, such as GPT Images, Gemini native image, and the async video models, keep their native fields; the documentation describes each contract separately instead of mapping them onto a single schema.

The catalog lists 14 image models (GPT Images, Gemini Image, Wan, Seedream) and 28 video models (Seedance, Wan, Kling, MiniMax, HappyHorse). Text models from GPT, Claude, Grok, and others are served over the OpenAI-compatible and Anthropic contracts.

Diagram: an application sends requests with one key to the Modelflare gateway, which serves six API families One key and one gateway in front of Chat Completions, Responses, Anthropic Messages, Images, Gemini native, and Video.

Endpoints

Family Endpoint Typical use
Chat Completions POST /v1/chat/completions OpenAI-compatible chat
Responses POST /v1/responses Codex, Grok Build
Anthropic Messages ANTHROPIC_BASE_URL=https://modelflare.dev Claude Code
Images POST /v1/images/generations GPT Images, Wan Image, Seedream
Gemini native POST /v1/models/{model}:generateContent Gemini image models (Gemini CLI builds the same path under /v1beta, served only on modelflare.dev)
Video POST /v1/videos, poll GET /v1/videos/{task_id}, download GET /v1/videos/{task_id}/content Seedance, Wan, Kling, MiniMax, HappyHorse
Models GET /v1/models Models your key can call

The base URL is https://modelflare.dev/v1. Anthropic clients use the site origin, https://modelflare.dev, without /v1, because the Claude Code CLI appends the Messages path itself. Gemini CLI reads GOOGLE_GEMINI_BASE_URL=https://modelflare.dev. China-facing EdgeOne traffic can use https://origin.modelflare.dev/v1. Choose the host before sending a request; an in-flight POST should not be redirected between hosts.

First request

Every request carries Authorization: Bearer YOUR_MODELFLARE_API_KEY.

curl https://modelflare.dev/v1/chat/completions \
  -H "Authorization: Bearer YOUR_MODELFLARE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.6-sol",
    "messages": [
      {"role": "user", "content": "Hello"}
    ]
  }'

Replace gpt-5.6-sol with any model your key can access. To list the models available to a key:

curl https://modelflare.dev/v1/models \
  -H "Authorization: Bearer YOUR_MODELFLARE_API_KEY"

Models and routing groups

Two identifiers decide whether a request succeeds. The model ID in the JSON body must match a public catalog ID exactly, without marketing names or date suffixes. The routing group bound to the key decides which provider channels can serve that model. Create the key in a non-default group that lists the model; if the group has no channel for it, the request fails even when the JSON is valid.

A Smart API Key removes the fixed group. It resolves a routing group on every request across all models the account can use, ordered by a routing strategy.

Pricing

Live prices, groups, and availability are published on Models & Pricing; the documentation does not keep a second price list. Language models are billed per million tokens with separate input and output rates. Image and video models are billed in the unit named on each model page: per token, per image, per call, or per second.

Resources on Hugging Face

The Modelflare organization hosts a landing Space, Modelflare/start; an SDK guide, Modelflare/openai-sdk; setup notes for coding agents, Modelflare/coding-agents; request recipes, Modelflare/api-recipes; and a dataset of example requests, Modelflare/api-examples.

Keys are created in the Modelflare console. Do not place a real key in a public Space or notebook. For business and enterprise inquiries, contact business@modelflare.dev.

← All articles