MiniMax models in short
| Model | Facts from the model card or docs |
|---|---|
| MiniMax M3 | About 428B total parameters and about 23B active. Native multimodal model with 1M context. License on Hugging Face: minimax-community |
| MiniMax M3.1 Flash | A smaller, lower-cost model in the M3 series. Listed on the MiniMax pricing page |
| MiniMax M2.7 | Listed by Groq as a preview model with a 196,608-token context window |
| MiniMax M2.5, M2.1, M2 | Earlier models. Weights are published on the MiniMaxAI Hugging Face page |
The M3 license is a MiniMax community license, not a standard open-source license. Read the license file on Hugging Face before commercial use.
Is the official MiniMax API free?
No. The MiniMax pricing page lists no free tier or free trial for the text models. It lists free allowances only for some image and video tasks.
| Model | Input per 1M tokens | Output per 1M tokens |
|---|---|---|
| MiniMax-M3 (up to 512K input) | $0.30 | $1.20 |
| MiniMax-M3 (above 512K input) | $0.60 | $2.40 |
| MiniMax-M3.1-Flash | $0.15 | $0.60 |
These prices are low, but they are not zero. For free use, you need a provider that serves MiniMax weights on its own free tier.
Free MiniMax routes today
The table shows every MiniMax row in today's free catalog. Each row lists the providers and the best free limit.
| Model | Free on | Context | Free limits (best offer) | Tools | Vision |
|---|---|---|---|---|---|
| MiniMax M3 | ElectronHub, Inferbase, ModelScope | 1M | 100 req/day Daily login: 200 credits + 50 for linked Alibaba Cloud; 24h expiry; shared; noncommercial experience | Yes | No |
| MiniMax: MiniMax M2.1 | ElectronHub | 200K | $0.25/week shared free credits | No | No |
| MiniMax M2.5 | ElectronHub, Septor Labs | 200K | $0.25/week shared recurring credit | Yes | No |
| MiniMax M2.7 | ElectronHub | 200K | $0.25/week shared recurring credit | Yes | No |
Live data from the FreeLLMAPI catalog (version 2026.11.19, checked 19 November 2026). Free limits change often. The catalog is checked every day.
- Shared credits: several routes use credit that all free users of that gateway share. A busy day can use it up.
- ModelScope: the free credits renew each day with a login, expire after 24 hours and are for noncommercial use. ModelScope also needs a linked Alibaba Cloud account.
- OpenRouter: on 20 November 2026, the OpenRouter model list had no free MiniMax variant.
- Groq: Groq lists MiniMax M2.7 as a preview model. We could not confirm the limits for it on the Groq free plan.
Call MiniMax through one free key
One shared free route stops often. FreeLLMAPI pools the MiniMax routes from all your providers behind one key. When one route returns a limit error, the router tries the next route for the same model. Do these steps:
- Install FreeLLMAPI and open the dashboard at
http://localhost:3001. - Add keys for two or more providers from the table above.
- Find the MiniMax model on the Models page. Copy its model id.
- Send requests with your unified key and that model id.
Request a MiniMax model through the gateway
curl http://localhost:3001/v1/chat/completions \
-H "Authorization: Bearer <unified-key>" \
-H "Content-Type: application/json" \
-d '{"model": "<minimax-model-id>", "messages": [{"role": "user", "content": "Hello"}]}'The same model on several providers shows as one entry with failover inside the group. The X-Routed-Via header shows the provider that answered.
Troubleshooting free MiniMax routes
| Problem | Cause | Fix |
|---|---|---|
| Error 429 or "quota exceeded" | The shared daily or weekly credit of the gateway is spent | Wait for the reset, or add a key for another route |
| Error 402 or "insufficient credit" | Your own credit on that gateway is spent | Use another route until your credit renews |
| Error on a long prompt | The free route has a small context, for example 8K tokens | Send a shorter prompt, or use a route with a larger context |
| Tool calls come back as plain text | The route does not support tool calling | Use a route with Yes in the Tools column |
| The answer is empty or cut off | The model spent its token budget on a thinking trace | Set a higher max_tokens value |
| ModelScope rejects the key | The account is not linked to Alibaba Cloud | Link the account, then log in again to get the daily credit |
With FreeLLMAPI, the router handles most of these errors for you. It puts the failed key on cooldown and tries the next route. The X-Fallback-Trail header lists each route that failed and the reason.
What MiniMax models are good for
- Long documents: MiniMax M3 has a 1M-token context on the model card. Free routes can give a much smaller context, for example 8K tokens. Check the Context column.
- Agents and coding: Some free MiniMax routes support tool calling and some do not. Check the Tools column, then test tool calls on your own tasks.
- Reasoning output: some routes return a thinking trace before the answer. Give the model enough
max_tokensfor both.
If a MiniMax route is full, similar free models can take the request. Read free coding LLM APIs and free long-context LLM APIs for other options.
Frequently asked questions
Is the MiniMax API free?
Not the official one. The MiniMax pricing page lists no free tier for text models. MiniMax M3 costs $0.30 per million input tokens and $1.20 per million output tokens at the standard rate.
Where can I use MiniMax M3 for free?
Several third-party providers serve MiniMax M3 on free credit, for example ModelScope and some gateways with shared credit. The table on this page shows the current routes and limits.
Is MiniMax M2 free?
The M2 series weights are on Hugging Face. MiniMax M2.5 and M2.7 also run on some free third-party routes. The table on this page shows which ones are free today.
Is MiniMax M3 open source?
The weights are published on Hugging Face under the minimax-community license. That is a custom license, so read its terms before commercial use.
Is there a free MiniMax model on OpenRouter?
Not on 20 November 2026. The OpenRouter model list had no free MiniMax variant on that date. Free variants on OpenRouter change often.
Related guides
- Best free coding model APIs
- Free LLM API with 1 million token context
- Free reasoning model API
- Free OpenAI-compatible API base URLs
- Free LLM API rate limits
Sources
- MiniMax platform: Pricing (checked 20 November 2026)
- Hugging Face: MiniMaxAI/MiniMax-M3 model card (checked 20 November 2026)
- Hugging Face: MiniMaxAI organization (checked 20 November 2026)
- GroqCloud docs: Supported models (checked 20 November 2026)
- OpenRouter: Models API (checked 20 November 2026)