{
  "data": null,
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/niches/image-generation.model-routing/",
    "json": "https://www.anchorterminal.com/niches/image-generation.model-routing/index.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/niches/image-generation.model-routing/index.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/niches/image-generation.model-routing/index.min.md"
  },
  "markdown": "No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.\n\n| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |\n| --- | --- | --- | --- | --- |\n| 0 | 11 | 45 | 14 | 7 |\n\n## What counts as done\n\nThe agent reads each candidate's schema and per-image price in one call, runs the same prompt on two models with one credential, and the chosen model accepts the same input shape.\n\nHow agents ask for it: which model is best for text; use the cheapest model for drafts; try this on FLUX and Nano Banana; use the new Seedream\n\n## Does the job, with a limit\n\nIt does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.\n\n- **Black Forest Labs FLUX API (Black Forest Labs Inc.)**: One key covers FLUX 3, FLUX.2 and Kontext with a cost field on each submit, but only BFL's models, and inputs differ (an images array on FLUX 3, input_image to input_image_8 on FLUX.2), as does the billing unit (per image or per megapixel).\n- **Bria API (Bria Artificial Intelligence Ltd.)**: Only Bria's own Fibo 2 and Fibo Lite are behind the key, with an OpenAPI bundle for schemas but no call that returns a model's price.\n- **fal image models (Features \u0026 Labels, Inc.)**: Each model's llms.txt gives schema and price in one fetch and one key runs all of them, but input fields differ by model (image_url, image_urls) and billing units differ (per image, per megapixel), so a payload doesn't carry across.\n- **FLORA (FLORA AND FAUNA FLOWERSHOP, INC.)**: GET /models returns each model's id, params, typical run time and base dollar price in one call on one key, but params is free-form per model (image_url on one, image_urls on another), so payloads don't carry across.\n- **Ideogram API (Ideogram AI, Inc.)**: One key reaches Ideogram 4.5, 4.0 and 3.0 plus GPT Image, Nano Banana and Z-Image, each with dry_run pricing, but fields differ per model path and acc-33 shows the spec changing without a changelog.\n- **Leonardo.Ai API (Leonardo Interactive Pty Ltd)**: One endpoint and key take some 40 image models (Lucid, Phoenix, FLUX, Ideogram, GPT Image, Nano Banana, Seedream, Recraft) selected by `model`, but each has its own parameter schema and no call returns a price before submitting.\n- **OpenAI Image API (OpenAI OpCo, LLC)**: One key reaches Sunburst and Flare with the same request shape, but only OpenAI's models, and price is per token with no per-image figure for the current models, so candidates can't be priced in one call.\n- **Recraft API (Recraft Inc.)**: One page lists every Recraft model's price and parameters on one token, but only Recraft's own models, parameter support differs by family, and the docs tell agents to check the table rather than rely on the API to reject a wrong parameter.\n- **Replicate image models (Replicate, LLC)**: One token runs every model with one request shape, but inputs differ per model: acc-27 hit format errors across three models and acc-32 saw seedream-4 hard-fail on sizes outside 1024 to 4096; official models bill per image, community per GPU-second.\n- **Seedream on BytePlus ModelArk (BytePlus Pte. Ltd.)**: One endpoint and key cover Seedream 5.5, 5.0 pro, flash and lite, 4.5 and 4.0 at published per-image prices, but only ByteDance's models, and size ranges and functions (layers, OCR) differ by version.\n- **Stability AI Image API (Stability AI Ltd)**: One key covers Ultra, Core, SD 3.5 and SDXL with fixed credit prices, but only Stability's models, and SDXL keeps a different v1 request shape from the v2beta services.\n\n## Couldn't check\n\nThe docs were closed to us, or silent. These carry no mark either way.\n\n- **Qwen-Image 2.1 (Model Studio) (Alibaba Cloud (Singapore) Private Limited)**: Model Studio pages couldn't be reached to check per-model schemas and prices on one key.\n\n- not this job: Adobe Firefly API (Adobe Inc.): Only Adobe's own Firefly Image 3, 4 and 5 and custom models, on two endpoint versions, with no price in any response or page.\n- not this job: Iris-3B (Speridlabs): A single local model with a CLI and Gradio demo; no API, catalogue or price to route between.\n- not this job: SurgePix API (SurgePix): No model catalogue, schema or price endpoint; the agent describes the asset and a skill picks for it.\n\n## What stops agents finishing\n\n- per-model input rules differ so a working payload fails on another model\n- API spec changes without a changelog\n- billing unit differs per model (per image, per megapixel, per token)\n\n## Around it\n\nAfter: image-generation.text-to-image, image-generation.project-asset-batch.\n\n## Where this comes from\n\nWeight 14 of 100 in Image generation: 7 of 37 accounts; a Claude plugin routing over 10 providers, Flux 2 against Nano Banana comparisons, Replicate models rejecting inputs others accept, spec drift on Ideogram\n\nAs data: https://www.anchorterminal.com/api/v1/use-cases/image-generation.model-routing.json, or the MCP tool get_use_case with id image-generation.model-routing. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-16",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.8",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "For agents",
        "url": "https://www.anchorterminal.com/agents/"
      },
      {
        "name": "Niches",
        "url": "https://www.anchorterminal.com/niches/"
      },
      {
        "name": "Choose the right model for the job and switch when a better one ships",
        "url": ""
      }
    ],
    "description": "Choose the right model for the job and switch when a better one ships: 0 companies fully do it, 11 with a limit, 45 carry the tag. Every listing in Image generation with the reason and a source.",
    "facts": null,
    "h1": "Choose the right model for the job and switch when a better one ships",
    "image": "https://www.anchorterminal.com/assets/og/default.png",
    "path": "/niches/image-generation.model-routing/",
    "published": "",
    "section": "agents",
    "title": "Choose the right model for the job and switch when a better one ships…",
    "toc": null,
    "updated": "2026-11-15",
    "url": "https://www.anchorterminal.com/niches/image-generation.model-routing/"
  },
  "tokens": {
    "markdown": 1450,
    "slim": 380
  },
  "version": 1
}
