Minimax Free API: Integration Guide, Pricing & Code

What “free” actually means for the MiniMax API

Search “minimax free api” and you’ll land on three different answers. MiniMax’s official docs mention free credits. OpenRouter lists an M2.5 model at $0 per token. Puter.js promises unlimited MiniMax access without API keys — but your users pay instead of you.

These aren’t competing claims. They’re three separate channels with different models, different limits, and fundamentally different definitions of “free.” Here’s what each one actually gets you:

Channel	What’s free	Models available	The catch
MiniMax official	Free credits on signup	All modalities (LLM + video + image + speech)	Credits run out; then pay-as-you-go
OpenRouter	M2.5 model at $0/token	MiniMax M2.5 only	Rate-limited; M3 and others are paid
Puter.js	Developer pays $0	M3, M2.7, M2.5, M2.1, M2-her	Your users see a payment gate

MiniMax free API access across three channels: official credits cover all modalities, OpenRouter limits to M2.5, and Puter.js shifts costs to users.

This article walks you through all three — with working code — so you can pick the path that fits your stack.

One disclosure before we start: we run Modellix, an aggregator that offers MiniMax models with pay-as-you-go pricing. We have a commercial interest. Every fact here about MiniMax’s own API comes from their public docs as of July 29, 2026. Where we couldn’t verify something, we say so.

Getting your free MiniMax API key (official path)

If you want direct access to MiniMax’s full model lineup — LLMs, video, image, and speech — the official path is where you start.

Step 1: Create an account. Go to platform.minimax.io and sign up. No credit card required for registration.

Step 2: Get your API key. Navigate to API Keys in the dashboard and create a new secret key. MiniMax gives new accounts free credits to test with — the exact amount varies and MiniMax doesn’t publish a fixed number, but expect enough for several hundred chat completions or a handful of video generations.

Step 3: Export your key. Drop this into your terminal or .env file:

export MINIMAX_API_KEY="eyJhbGciOi..."   # your key here

MiniMax’s API is compatible with both OpenAI and Anthropic SDKs. The base URLs:

OpenAI-compatible:  https://api.minimax.io/v1
Anthropic-compatible: https://api.minimax.io/anthropic

Rate limits on free credits are unpublished, but pay-as-you-go accounts get higher throughput. If you hit 429 Too Many Requests, you’ve either exhausted your free credits or hit the rate ceiling — top up or slow down.

Your first MiniMax API call — direct integration code

Most tutorials route through OpenRouter or a JavaScript SDK. Here’s how to call api.minimax.io directly — no wrapper, no third-party hop.

curl

curl https://api.minimax.io/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $MINIMAX_API_KEY" \
  -d '{
    "model": "minimax-m2.5",
    "messages": [{"role": "user", "content": "Explain bloom filters in one paragraph."}],
    "max_tokens": 200
  }'

Python

import os, requests

resp = requests.post(
    "https://api.minimax.io/v1/chat/completions",
    headers={
        "Authorization": f"Bearer {os.environ['MINIMAX_API_KEY']}",
        "Content-Type": "application/json"
    },
    json={
        "model": "minimax-m2.5",
        "messages": [{"role": "user", "content": "Explain bloom filters in one paragraph."}],
        "max_tokens": 200
    }
)
print(resp.json()["choices"][0]["message"]["content"])

Node.js

const resp = await fetch("https://api.minimax.io/v1/chat/completions", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.MINIMAX_API_KEY}`,
    "Content-Type": "application/json"
  },
  body: JSON.stringify({
    model: "minimax-m2.5",
    messages: [{ role: "user", content: "Explain bloom filters in one paragraph." }],
    max_tokens: 200
  })
});
const data = await resp.json();
console.log(data.choices[0].message.content);

Model names you can use: minimax-m2.5, minimax-m3, minimax-m2.7, minimax-m2.1. MiniMax’s model list evolves quickly — check their OpenAI-compatible API docs for the current roster.

Common errors:

• 401 Unauthorized — your API key is missing, expired, or malformed. Double-check the Authorization: Bearer header.
• 429 Too Many Requests — rate limit hit. Wait and retry, or top up your balance.
• model not found — you’re using a model name that doesn’t exist or isn’t available on your account tier.

Streaming? Add "stream": true to the request body. Same endpoint, same auth.

Free MiniMax through OpenRouter (zero-cost model access)

OpenRouter lists minimax/minimax-m2.5:free at $0 per token. It’s genuinely free — OpenRouter subsidizes it — but only for that one model, and with rate limits.

Step 1: Get an OpenRouter API key. Sign up at openrouter.ai (free, no payment required), create an API key, and you’re done.

Step 2: Call MiniMax M2.5 through OpenRouter’s endpoint:

curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "minimax/minimax-m2.5:free",
    "messages": [{"role": "user", "content": "Write a Python decorator that retries a function on exception."}]
  }'

The model ID is minimax/minimax-m2.5:free — the :free suffix is mandatory. Without it, you get the paid version.

What’s not free on OpenRouter: MiniMax M3, M2.7, M2.1, Hailuo video models, and speech models all have per-token pricing. Only M2.5 carries the free tag. If your use case needs M3’s 1M-token context window or Hailuo’s video generation, this path won’t cover it.

Rate limits: OpenRouter doesn’t publish a fixed RPM for the free tier, but expect throttling under sustained load. For production workloads, the paid models on OpenRouter or MiniMax’s own API are the realistic options.

When OpenRouter makes sense: You want free LLM access, don’t need video/image modalities, and are fine with a single model (M2.5). The zero-cost entry point is unbeatable for prototyping.

Free MiniMax through Puter.js (user-pays model)

Puter.js takes a different approach: you, the developer, pay nothing. Your app’s users cover the AI costs through Puter’s user-pays infrastructure. This model works if you’re building a web application where users expect to pay for AI features.

Step 1: Add Puter.js to your project. Drop the CDN script into your HTML or install the npm package:

<script src="https://js.puter.com/v2/"></script>

Step 2: Call any MiniMax model. No API key, no billing setup:

<script>
  puter.ai.chat(
    "Build a CLI tool in Go that watches a directory and runs a configurable build command.",
    { model: "minimax/minimax-m3" }
  ).then(response => {
    console.log(response);
  });
</script>

Puter.js supports M3, M2.7, M2.5, M2.1, and M2-her, plus Highspeed variants of M2.7, M2.5, and M2.1. This is the widest free model selection of any channel — but it comes with the user-facing payment gate.

The trade-off: Your app stays free to run, but your users hit a payment prompt when they exceed usage limits. For developer tools, internal dashboards, or apps where users expect to pay for compute, this model works. For a public-facing free service, it’s friction your users didn’t sign up for.

Beyond LLMs — free access to MiniMax video and image APIs

MiniMax isn’t just a text company. Their Hailuo line generates video, their image-01 model produces images up to 1440×2560, and their speech models handle TTS. The free credits on a new MiniMax account work across all of these — not just LLMs.

Video generation (Hailuo)

Hailuo video models consume credits fast. A single 6-second generation at 768p costs 1.0 point; at 1080p it’s 2.0 points. Your free credits will cover a handful of generations — enough to evaluate quality, not enough for a production pipeline.

The endpoint is the same base URL but uses MiniMax’s native video API format. Check their API overview for the current video generation spec — the endpoint structure changes between model versions.

On OpenRouter, Hailuo 2.3 is listed as a paid model — there’s no free video tier. On Puter.js, MiniMax video models aren’t currently available through the chat interface.

Image generation

MiniMax image-01 generates images from text prompts through the same OpenAI-compatible endpoint. Your free credits cover both text and image calls from the same balance. On Modellix, image-01 starts at $0.004 per image — the cheapest model across all 12 suppliers on the platform — but that’s paid, not free.

MiniMax-01: the open-weight option

MiniMax-01 is a 456-billion-parameter model (45.9B active) with a 4-million-token context window, released as open weights. It’s available on Hugging Face for self-hosting — genuinely free if you have the GPU capacity. OpenRouter lists it as a paid model. MiniMax-01 is a combined text + vision model (MiniMax-Text-01 + MiniMax-VL-01), making it relevant for image understanding pipelines. If your team has the infrastructure, self-hosting MiniMax-01 is the only path to truly unlimited free usage — but the hardware cost dwarfs any API bill.

What happens when “free” runs out — pricing comparison

Every free path has an exit. Here’s what you pay when you cross the line:

Channel	Model	Free limit	Paid pricing
MiniMax official	M2.5 text	~500+ completions	$0.30/$1.20 per 1M input/output tokens
MiniMax official	Hailuo 2.3 video	~5-10 generations	$1,000/month prepaid package (3,760 points)
OpenRouter	M2.5 text	Rate-limited, $0/token	M3: market-rate per token
Puter.js	All models	Usage-based per user	Users pay per request

For MiniMax’s video API specifically, the jump from free credits to paid is steep — the cheapest production package is $1,000 per month for 3,760 video points, with a 30-day expiry. We covered this in detail in our MiniMax pricing breakdown, and our Hailuo API guide compares Hailuo 02 vs 2.3 with native 1080p benchmarks.

If you need MiniMax models — especially video — without the $1,000 monthly commitment, Modellix offers MiniMax Hailuo models with per-second billing, no prepaid packages, and a single API key that also covers 210+ other image and video models. New accounts get a free $10–$30 credit, no credit card required. We don’t carry MiniMax’s LLM models (M3, M2.5) — for those, MiniMax’s own API or OpenRouter are your options.

For LLM-only usage, MiniMax’s Token Plans at $20–$120/month give you bundled credits that don’t expire as aggressively as the video packages. The official pay-as-you-go rates for M2.5 ($0.30/$1.20 per 1M tokens) are competitive with other frontier models, and the free credits on signup give you room to evaluate before committing.

Which free access path should you use?

The answer depends on which models you need and how your app is built:

Use MiniMax official free credits if you need the full model lineup — LLMs, video, image, and speech — and want direct API access without routing through a third party. Best for evaluating MiniMax’s entire platform before picking a paid plan.

Use OpenRouter’s free M2.5 if you only need text generation, want zero-cost prototyping, and don’t mind being limited to one model. The :free suffix is the cleanest zero-cost LLM path available today — no credits to exhaust, no payment gate for users.

Use Puter.js if you’re building a web app where users expect to pay for AI features, and you want to ship without a billing infrastructure. You get the widest free model selection (M3, M2.7, M2.5, M2.1, M2-her) but your users carry the cost.

Consider an aggregator like Modellix if you need MiniMax video at production scale without the $1,000/month prepaid commitment, or if you’re routing between multiple model families (Hailuo for one type of shot, something else for another) and want one API key and one invoice. Our MiniMax model collection has live pricing and a playground to test before you write code.

The free credits on any path are a trial — not a business model. Pick the channel whose paid tier matches your scale, and use the free access to validate that MiniMax’s models do what you need.

---

All pricing and model availability verified against MiniMax’s official docs, OpenRouter, and Puter.js on July 29, 2026. MiniMax’s pricing, model names, and free credit amounts change without notice — verify against the linked pages. Modellix offers MiniMax’s Hailuo models with pay-as-you-go billing; we do not offer MiniMax’s LLM models. We have a commercial interest in the aggregator comparison.