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.

LongCat 2.0 vs MiniMax M3: API price, context & features · api14

Meituan: LongCat 2.0 vs MiniMax: MiniMax M3

A side-by-side comparison of Meituan: LongCat 2.0 ($1.05 / 1M tokens) and MiniMax: MiniMax M3 ($1.05 / 1M tokens) using current USD prices and capabilities from the api14 catalog. Both run through the same OpenAI-compatible API.

Specifications

	Meituan: LongCat 2.0	MiniMax: MiniMax M3
Provider	Meituan	MiniMax
Input	$1.05 / 1M tokens	$1.05 / 1M tokens
Output	$4.20 / 1M tokens	$4.20 / 1M tokens
Cache read	$0.02 / 1M tokens	$0.21 / 1M tokens
Context window	1,048,756	1,048,576
Max output	262,144	512,000
Input	text	text, image, video
Output	text	text
Tool calling
Structured output
Reasoning control
Added	Jul 20, 2026	May 31, 2026

Key differences

Both have about the same input price.
Only MiniMax: MiniMax M3 supports structured JSON output.
A request with 1,000 input and 500 output tokens costs $0.00315 with Meituan: LongCat 2.0 and $0.00315 with MiniMax: MiniMax M3.

In depth

Meituan: LongCat 2.0 is a text model from Meituan, released on 20 July 2026. MiniMax: MiniMax M3 comes from MiniMax and arrived earlier, on 31 May 2026. Both are listed on api14 under the text category, and both are served through the same OpenAI-compatible interface, so the choice between them is mostly about what you need to send in and what you need to get back, not about how you integrate.

The clearest split is in input modalities. LongCat 2.0 accepts text only and returns text. MiniMax M3 accepts text, images and video as input, and returns text. If your prompt includes a screenshot, a scanned page, a chart or a video clip, M3 is the one that can take it directly, while LongCat 2.0 would need you to convert that content into text first. The context windows are effectively the same size, both a little over one million tokens, so neither model gives you a meaningful advantage on how much you can fit into a single request. On parameters the two overlap heavily, with temperature, top_p, top_k, min_p, penalties, seed, stop, tools, tool_choice and reasoning controls available on both. M3 adds logprobs, top_logprobs, response_format and structured_outputs, which LongCat 2.0 does not list.

For day to day work, pick based on the shape of the input and the shape of the output you need. If you are building a pipeline that must return strict JSON or another schema you can validate, M3 is the safer default because structured outputs and response_format are supported, and you can lean on that instead of parsing free text and hoping. If you want token level confidence signals for classification, routing or scoring, logprobs and top_logprobs on M3 give you that, and LongCat 2.0 does not offer them. LongCat 2.0 remains a reasonable pick for plain text in, text out tasks where you do not need schema enforcement or log probabilities, and where you would rather keep the request surface simple. Both models support tool calling and reasoning controls, and the qualitative pricing comparison puts them at about the same price on both input and output, so cost is not the deciding factor here.

Calling either one through api14 is the same code path. You point your OpenAI client at the api14 base URL, send your key, and set the model field. To use Meituan's model, set the model id to "meituan/longcat-2.0". To use MiniMax's model, set it to "minimax/minimax-m3". Everything else in the request body follows the OpenAI chat completions shape, so switching between the two is a one line change in the model id, plus whatever extra fields the target model supports. If you rely on structured_outputs or logprobs, keep those in the request only when you are calling M3, since LongCat 2.0 does not list them.

There is one case where LongCat 2.0 is still the better pick even though M3 covers more ground. If your task is purely text and you want the smallest set of request fields to reason about, or you are already standardized on a text only contract and do not want image or video handling anywhere near your code path, LongCat 2.0 keeps that contract clean. The two are about the same price, so you are not paying a premium for the narrower model, and you are not giving up context length either.

Switch with one string

Both models use the same endpoint and request shape. Change the model value from "meituan/longcat-2.0" to "minimax/minimax-m3" and prompts, streaming and tools keep working.

Meituan: LongCat 2.0 details & playground →MiniMax: MiniMax M3 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": "minimax/minimax-m3",
  "messages": [
    {
      "role": "user",
      "content": "Explain vector databases in two sentences."
    }
  ]
});
console.log(result.choices[0].message.content);

Meituan: LongCat 2.0 vs MiniMax: MiniMax M3

Specifications

Key differences

In depth

Switch with one string

More comparisons