EdgeeOne gateway for cheaper, faster, unstoppable coding agents“We often use minimax models connected to our agents, the performance is amazing, really close to Opus. And the price is really interesting.”
LLM APIMaker says so · source ↗MiniMax
API for MiniMax's own models — open-weight M-series text models, speech, image and video — with OpenAI- and Anthropic-compatible endpoints.
Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗
MiniMax is an AI lab that sells API access to its own models through its international platform. The current text models are MiniMax-M3, with a 1M-token context window and text, image and video input, and the 200k-context MiniMax-M2.7 with a faster, pricier highspeed variant; M3 and M2.7 are also published as open weights on Hugging Face. The same account covers text-to-speech, image generation and the Hailuo and H3 video models.
There is no MiniMax-specific SDK to learn. The API speaks the Anthropic Messages format at https://api.minimax.io/anthropic, which MiniMax recommends, and the OpenAI format at https://api.minimax.io/v1, so you use the official Anthropic or OpenAI SDKs, or agent tools such as Claude Code, by changing the base URL and key. Reasoning comes back in a separate field from the answer. Pay-as-you-go usage is deducted from a prepaid balance.
Useful for a small team: low per-token prices on M3 and M2.7, automatic caching that bills repeated prompt prefixes at a lower rate without code changes, and per-model limits in requests and tokens per minute (for M3, 200 requests and 10M tokens per minute). For use inside coding tools there is also M Plan, a monthly subscription that works in Claude Code, Codex, Cursor and OpenCode as well as MiniMax's own agent app.
The main limits: the newest text model, MiniMax-M3.1-Flash-Preview, is only available through M Plan and MiniMax Code, not pay-as-you-go. M3 input beyond 512k tokens is billed at a higher rate, and the paid music and lyrics APIs closed to new users on August 20, 2026. Self-hosting the open-weight models needs multi-GPU servers.
Where it fits
Who uses it
6 makers' products, each linked to the source that shows it, and 2 open-source projects that declare it in their code.
EdgeeOne gateway for cheaper, faster, unstoppable coding agents“MiniMax H3 models were amazing at creating sprite animations!”
LLM APIMaker says so · source ↗“MiniMax powers our Voice Clone feature. Its cloning technology delivers remarkably accurate voice replication with minimal audio input, which was the key factor over alternatives we tested.”
LLM APIMaker says so · source ↗“Zoer.ai utilizes MiniMax’s advanced reasoning capabilities to ensure high precision in complex task planning and logical orchestration.”
LLM APIMaker says so · source ↗“Cheers to MiniMax for making our characters chat like they’re real!”
LLM APIMaker says so · source ↗
TelnyxCommunications APIs for SMS/MMS, voice, phone numbers and SIP trunking on Telnyx's own n…LLM APISubprocessor list · source ↗Open source: a project that declares MiniMax as a dependency in its public code — verifiable, but not necessarily a live product.
What makers say
5 makers on why they use MiniMax, in their own words on Product Hunt.
We often use minimax models connected to our agents, the performance is amazing, really close to Opus. And the price is really interesting.
EdgeeSep 2026 ↗MiniMax powers our Voice Clone feature. Its cloning technology delivers remarkably accurate voice replication with minimal audio input, which was the key factor over alternatives we tested.
Zoer.ai utilizes MiniMax’s advanced reasoning capabilities to ensure high precision in complex task planning and logical orchestration.
Cheers to MiniMax for making our characters chat like they’re real!
MiniMax H3 models were amazing at creating sprite animations!
Loved and watch-outs
Themes that recur in makers' words and Hacker News comments, each linked to what it summarises, with how Product Hunt tags its reviews.
Reliability and open issues
Incidents on its status page50+ in the last 90 days, 50 major · 50+ in the last year
- major Elevated errors on Large Language Models (LLM) Oct 2026
- major Elevated errors on Large Language Models (LLM) Oct 2026
- major Elevated errors on Large Language Models (LLM) Oct 2026
Alternatives to MiniMax
All alternatives by situation →
OpenAI API
Claude
Gemini API
Groq
Mistral AI
DeepSeekOn Product Hunt, people weigh it against: OpenAI API, Mistral AI, Claude, Grok API, Gemini API.
Questions makers ask about MiniMax
Can I use the OpenAI or Anthropic SDK?
Yes. Set the base URL to https://api.minimax.io/anthropic for the Anthropic SDK (MiniMax's recommended option) or https://api.minimax.io/v1 for the OpenAI SDK, and use a MiniMax API key. On the OpenAI format, reasoning is returned in reasoning_content, separate from the answer. source ↗
Where is my data stored?
The international platform's privacy policy says personal data is stored in a data center in the United States, and that MiniMax doesn't use personal data from your inputs for training to profile or target consumers. source ↗
What are the rate limits?
Limits are set per model in requests and tokens per minute — 200 RPM and 10M TPM for MiniMax-M3, and 500 RPM and 20M TPM for MiniMax-M2.7 and its highspeed variant. source ↗
Do I need to set up prompt caching?
No. Automatic caching detects repeated context and bills cache hits at a discounted rate, with no charge for writes. If you use explicit cache_control on the Anthropic-compatible API instead, the first write to the cache costs extra. source ↗
Can I call MiniMax-M3.1-Flash-Preview with a pay-as-you-go key?
Not for now. MiniMax says the model is available only through an M Plan subscription and its MiniMax Code app; pay-as-you-go keys cover M3 and M2.7. source ↗
Can I still use the music generation API?
Only if you were already a paying user. From August 20, 2026, the paid music and lyrics APIs are closed to new users and the free music APIs are discontinued; MiniMax points to its consumer audio app or the open-weight Music 3 model instead. source ↗
What is M Plan?
A monthly subscription in three tiers (Go, Explore, Build) with a separate subscription key, for using MiniMax models in MiniMax Code or in tools such as Claude Code, Codex, Cursor and OpenCode, instead of paying per token. source ↗
Is MiniMax free?
No permanent free tier; paid use starts at Pay per token. source ↗
Is MiniMax open source or self-hostable?
Not open source, and hosted only.
Can AI coding agents work with MiniMax?
No llms.txt, official MCP server or CLI found yet.
Who uses MiniMax?
6 makers' products we track, each with a source, and 2 open-source projects declare it in their code. source ↗