Can Meta Muse Wonder generate images? It returns text; use /v1/images
Muse Wonder is listed on OpenRouter as a chat-completions model with text responses. It is not in Sume's image catalog; pick an image row instead.
No, not as far as the listing I could read shows. OpenRouter's page for meta/muse-wonder documents a POST /api/v1/chat/completions call: you send messages and get message content and token usage back, with parameters for tools, reasoning and response_format. It describes no image output, and Sume's image catalog does not list a Muse Wonder row.
What the listing shows
I could not read Meta's own model card (the file would not parse in my fetcher), so the facts below come from the OpenRouter page only, and I do not state a release date or price here.
| Item | What the page shows |
|---|---|
| Endpoint | POST /api/v1/chat/completions |
| Response | Message content plus token usage |
| Parameters | stream, max_tokens, temperature, top_p, top_k, reasoning, reasoning_effort, tools, tool_choice, response_format, repetition_penalty |
| Image output | Not described |
What to call for pictures
Muse Image, a separate Meta product, is covered in a separate post on what Sume lists. For images today, Sume's POST /v1/images takes a catalog id. ChatGPT Image 2.5 is the row with the widest edit controls: mask_url, background and up to 16 references. The sample asks for a transparent PNG asset.
Replace the model id with any row from GET /v1/images/models; unknown ids return 404 model_not_found.
curl -s https://api.sume.com/v1/images \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"openai/gpt-image-2.5","prompt":"flat vector fox mascot","background":"transparent","output_format":"png"}'Check before you build
- Search the catalog for the exact model name before you pin an id.
- Do not assume that a model with an image-capable sibling can draw.
- Treat a text-only model as a prompt writer at best, and test the result.
Sources
Related posts
More in Models
- Decision-1 GlobalStandard or DataZoneStandard for a Sume gate?
Microsoft lists both deployment types for Decision-1. DataZoneStandard keeps processing in a data zone; GlobalStandard may vary more in latency. Benchmark.
- Mistral Large 4 function calling: validate arguments against Sume
Mistral Large 4 (Public Preview) lists function calling and structured outputs. Validate its arguments with Sume's tools_schema before any paid call.
- Muse Wonder reviews a video: send Sume inspect stills and transcript
Muse Wonder takes video and audio in but replies only in text. To have it review a Sume render, use video_inspect: 8 stills by default, 24 at most.
- Nano Banana 2.1 thinking levels and search grounding: not on Sume
fal documents thinking levels, web-search grounding and a system prompt for Nano Banana 2.1. Sume's /v1/images row lists none and returns 400 for them.
Written by Sume