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.

Codestral Embed 2505 vs Qwen3 Embedding 4B: API price, context & features · api14

Mistral: Codestral Embed 2505 vs Qwen: Qwen3 Embedding 4B

A side-by-side comparison of Mistral: Codestral Embed 2505 ($0.52 / 1M tokens) and Qwen: Qwen3 Embedding 4B ($0.07 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Mistral: Codestral Embed 2505	Qwen: Qwen3 Embedding 4B
Provider	Mistral	Qwen
Input	$0.52 / 1M tokens	$0.07 / 1M tokens
Context window	8,192	32,768
Max output	6,553	29,491
Input	text	text
Output	embeddings	embeddings
Tool calling
Structured output
Reasoning control
Added	Oct 30, 2025	Oct 28, 2025

Key differences

Qwen: Qwen3 Embedding 4B is 87% cheaper on input tokens.
Qwen: Qwen3 Embedding 4B reads 4.0× more context (32,768 tokens).

In depth

Mistral: Codestral Embed 2505 is an embeddings model from Mistral, released on 30 October 2025. Qwen: Qwen3 Embedding 4B is an embeddings model from Qwen, released on 28 October 2025. Both sit in the embeddings category, so neither one generates chat completions; you send text in and get vectors back. That shared purpose is the starting point for the comparison, because the choice between them comes down to context size and price rather than to different kinds of output.

The clearest difference is the context window. Codestral Embed 2505 accepts up to 8,192 tokens of input, while Qwen3 Embedding 4B accepts up to 32,768 tokens, four times as much. Both take text as input and both return embeddings as output, so there is no modality gap to weigh: neither model handles images, audio or anything other than text. Neither lists any supported parameters, so you should not expect to tune sampling or similar options on either call. On price, Qwen3 Embedding 4B is the cheaper of the two for input, and the comparison data does not distinguish them on output cost.

For practical work, the decision usually follows the length of the documents you need to embed. If you are embedding short passages, individual functions, docstrings or small chunks of source code, Codestral Embed 2505 fits comfortably inside its 8,192 token limit and its code-oriented background makes it a reasonable default for repository search and code retrieval. If your inputs are long files, whole modules, lengthy documentation pages or large chunks you would rather not split, Qwen3 Embedding 4B is the better fit because its 32,768 token window lets you keep more of a document intact in a single request, and it is also the cheaper option on input, which matters when you are embedding a large corpus.

Calling either model through api14 works the same way, because the API is OpenAI-compatible. You point your existing client at the api14 base URL, send your text to the embeddings endpoint, and change only the model id: "mistralai/codestral-embed-2505" for the Mistral model or "qwen/qwen3-embedding-4b" for the Qwen model. Your request shape, authentication and response parsing stay identical, so you can switch between them or run both side by side without rewriting integration code. The live price table on the page shows the current rates for each.

There are still cases where Codestral Embed 2505 is the better pick despite the smaller window. If your pipeline already chunks text into short segments, the context difference stops mattering, and a code-focused embedding model is a sensible choice for indexing source code and matching queries against it. In that situation the 8,192 token limit is not a constraint, and you can choose based on how well each model represents your particular content rather than on window size alone.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "mistralai/codestral-embed-2505" to "qwen/qwen3-embedding-4b" and prompts, streaming and tools keep working.

Mistral: Codestral Embed 2505 details & playground →Qwen: Qwen3 Embedding 4B 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.embeddings.create({
  "model": "qwen/qwen3-embedding-4b",
  "input": [
    "The quick brown fox",
    "Lorem ipsum"
  ]
});
console.log(result.data.map((d) => d.embedding.length));

Mistral: Codestral Embed 2505 vs Qwen: Qwen3 Embedding 4B

Specifications

Key differences

In depth

Switch with one string

More comparisons