We use cookies. Cookies and browser storage keep you signed in and remember your preferences. With your permission we also measure how the site is used. You can any time. See our cookie policy.

New here? See how it works A 1-minute tour of the playground, models and your API key.

Muse Wonder vs Extract: API price, context & features · api14

Meta: Muse Wonder vs ScaleDown: Extract

A side-by-side comparison of Meta: Muse Wonder ($14.00 / 1M tokens) and ScaleDown: Extract ($0.14 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Meta: Muse Wonder	ScaleDown: Extract
Provider	Meta	ScaleDown
Input	$14.00 / 1M tokens	$0.14 / 1M tokens
Output	$70.00 / 1M tokens	-
Cache read	$0.70 / 1M tokens	-
Web search	$8.75 / 1K searches	-
Context window	1,048,576	1,000,000
Max output	943,718	900,000
Input	text, image, video, file, audio	text, image, file
Output	text	text
Tool calling
Structured output
Reasoning control
Added	Oct 12, 2026	Oct 13, 2026

Key differences

ScaleDown: Extract is 99% cheaper on input tokens.
Only Meta: Muse Wonder supports tool calling.
Only Meta: Muse Wonder supports structured JSON output.
Only Meta: Muse Wonder supports reasoning effort control.
A request with 1,000 input and 500 output tokens costs $0.05 with Meta: Muse Wonder and $0.000140 with ScaleDown: Extract.

In depth

Meta: Muse Wonder is a text model from Meta, listed on api14 under the model id meta/muse-wonder. ScaleDown: Extract is a text model from ScaleDown, listed under scaledown/extract. Both are text-in, text-out models in the sense that their output modality is text only, and both were released within a day of each other in October 2026, so neither is a legacy option you have to work around. The interesting differences are in what they accept as input and what you can ask them to do.

Muse Wonder takes text, image, video, file and audio as input, while Extract takes text, image and file. So if your pipeline involves audio or video, Muse Wonder is the one that can look at that material directly. Context length is close enough that it should not drive your decision: Muse Wonder lists 1,048,576 tokens and Extract lists 1,000,000, and the qualitative comparison treats the context difference as not meaningful. The bigger split is in supported parameters. Muse Wonder exposes tools and tool_choice, structured_outputs and response_format, and the reasoning family (reasoning, reasoning_effort, include_reasoning), plus the usual sampling controls like temperature, top_p, top_k and repetition_penalty. Extract supports only max_tokens. That means anything involving function calling, enforced JSON schemas or explicit reasoning control has to go to Muse Wonder, because Extract gives you no lever for it.

For day to day work, pick based on what the task actually needs. If you are extracting fields from documents, images or plain text and you just want the answer back, Extract is the cheaper input option and its narrow parameter surface is not a handicap, since you were only going to set max_tokens anyway. If you need the model to call tools, return a schema-conformant object, or spend deliberate reasoning effort on a hard problem, Muse Wonder is the only one of the two that supports it. The same applies if your input includes audio or video, which Extract does not accept. For long mixed-media analysis where you also want structured output at the end, Muse Wonder covers the whole path in one call.

Calling either one through api14 works the same way, because the API is OpenAI-compatible. You point your existing client at the api14 base URL, send your usual chat completions request, and change only the model field: meta/muse-wonder or scaledown/extract. Your message structure, your auth header and your response parsing stay as they are. The one thing to watch is parameters: if you switch from Muse Wonder to Extract, drop everything except max_tokens, since Extract does not list support for temperature, tools or response_format. Going the other direction, you can start using those fields immediately.

There is still a case for Extract even when Muse Wonder is available. When the job is a straightforward text, image or file extraction and you do not need tools, structured output or reasoning controls, Extract is the cheaper input model and there is no reason to pay for capability you will not call. Reach for Muse Wonder when the task needs more than a single pass of reading and answering.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "meta/muse-wonder" to "scaledown/extract" and prompts, streaming and tools keep working.

Meta: Muse Wonder details & playground →ScaleDown: Extract details & playground →

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api14.com/api/v1",
  apiKey: process.env.API14_API_KEY,
});

const result = await client.chat.completions.create({
  "model": "scaledown/extract",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

Meta: Muse Wonder vs ScaleDown: Extract

Specifications

Key differences

In depth

Switch with one string

More comparisons