MiniMax-M3

MiniMax's top model, succeeding M2.7: a 428B open-weight MoE activating roughly 23B per token across 256 experts with sparse attention, trained natively on text, image and video.

Params: 428B, 23B active
Context: 1M tokens
Released: Jun 2026
License: MiniMax Community License

Visit MiniMax Compare vs other LLMs →

Hosted API price

List price via MiniMax · per 1M tokens

USD

Input: $0.30 /1M
Output: $1.20 /1M
Cheapest: $0.20 / $0.90 via EcoHash
Hosted by: 8 providers

Hosted API pricing

How we price

Order

Providers are listed cheapest input rate first, then by output rate, a tie going to the creator's own listing. A provider that lists the model without a published rate sits at the end.

Prices

Base-tier, on-demand rates in USD per 1M tokens, converted at ECB reference rates where a provider publishes in another currency. Cached-input and batch rates show only where a provider publishes them. "Cheapest" marks the one provider with the lowest input and output pair; a tie wears no badge.

Cost

Input rate times the input tokens plus output rate times the output tokens, at the monthly volume set above the table. Cached-input, batch, long-context and reasoning-token billing are not modelled.

Transparency and funding

Affiliates: Affiliate links are marked. We may earn a commission if you click them, but commissions never affect the order.

Every provider serving MiniMax-M3, cheapest input first. Set your monthly volume to see what each would bill.

Input M Output M

Every provider serving MiniMax-M3, with its price per 1M input and output tokens and what the volume above would cost
Provider	Input /1M	Output /1M	Cached input /1M	Cost at 10M in + 2M out	Link
EcoHash Cheapest	$0.20	$0.90	$0.05	$3.80	Visit website
EmpirioLabs AI	$0.225	$0.90	$0.045	$4.05	Visit website
MiniMax Creator	$0.30	$1.20	$0.06	$5.40	Visit website
GPUhub	$0.30	$1.20	$0.03	$5.40	Visit website
Novita	$0.30	$1.20	$0.06	$5.40	Visit website
Together AI	$0.30	$1.20	$0.06	$5.40	Visit website
Azure	$0.33	$1.32	$0.066	$5.94	Visit website
Lyceum	$0.40	$2.00	$0.10	$8.00	Visit website

Heads up: Base-tier, on-demand rates per 1M tokens; cached, batch and long-context tiers excluded. A provider may serve a shorter context or a quantized build than the creator's release. Verify before provisioning. More on how we price.

Estimated cost to self-host

MiniMax-M3 needs about 241 GB of GPU memory at 4-bit with 32K context. The cheapest rental that fits is 8× RTX A6000 at about $3,398 a month.

That costs the same as roughly 8B tokens a month on MiniMax's API. Self-hosting is more expensive below that volume.

Single node ≤ 8

What it costs to self-host MiniMax-M3 at each precision: the memory it needs, the cheapest rental that holds it, that rental's month and the API volume it pays off at
Precision	Memory	Cheapest fit (32K context)(Full 1M context)	Cost /mo	Break-even vs API
4-bit INT4 / FP4	241 GB299 GB	8× RTX A6000	$3,398$5,645	8B tokens /mo13B tokens /mo
8-bit FP8 / INT8	434 GB553 GB	8× A100	$10,714$10,714	24B tokens /mo24B tokens /mo
16-bit FP16 / BF16	862 GB981 GB	4× B300	$27,360$27,360	61B tokens /mo61B tokens /mo

Estimates based on median on-demand rates for Nvidia GPUs. Memory is weights plus KV cache for one request, using FP8 KV cache where supported and FP16/BF16 otherwise. Break-even assumes a 5:1 input-to-output ratio. No guarantee of runtime support, usable performance, or that a matching quantized build exists. Pricing methodology.

Similarly priced models

The models nearest MiniMax-M3 by blended rate, each at its own cheapest provider.

Models priced nearest MiniMax-M3 per 1M tokens, each at its cheapest provider
Model	Blended /1M	Input /1M	Output /1M	Context	Cutoff	vs MiniMax-M3
Qwen3-235B-A22B Instruct (2507) Alibaba Cloud	$0.27	$0.20	$0.60	262K		−16%
Qwen3-VL-30B-A3B Instruct Alibaba Cloud	$0.28	$0.20	$0.70	262K		−11%
Muse Glimmer 30B Meta	$0.30	$0.20	$0.80	131K	Jan 2026	−5%
Qwen2.5-7B Instruct Alibaba Cloud	$0.30	$0.30	$0.30	33K		−5%
Qwen3.5-397B-A17B Alibaba Cloud	$0.32	$0.172	$1.032	262K		0%
MiniMax-M3 This model MiniMax	$0.32	$0.20	$0.90	1M
Qwen-MT-Plus Alibaba Cloud	$0.33	$0.25	$0.75	16K		+5%
Llama 4 Maverick Meta	$0.35	$0.25	$0.87	1M	Aug 2024	+12%
GPT-5.6 Luna OpenAI	$0.37	$0.20	$1.20	1M	Feb 2026	+16%
GPT-5.4 Nano OpenAI	$0.38	$0.20	$1.25	400K	Aug 2025	+18%
Codestral Mistral	$0.40	$0.30	$0.90	128K		+26%

USD per 1M tokens at each model's cheapest listed provider. Blended is the cost of 10M input plus 2M output tokens, spread over the 12M.

Lyceum

NVIDIA H200 from $2.20/hr, B200 from $3.40/hr, B300 from $3.80/hr

Per-second billing with no minimum commitment or vendor lock-in

Scale to 8,000 GPUs over InfiniBand for distributed AI training

View Pricing

Capabilities

What MiniMax-M3 accepts and can do, as published by MiniMax.

Text Accepts and generates natural-language text.
Image Accepts images as input alongside text.
Video Accepts video as input.
Reasoning Thinks before it answers, always on or as a switchable mode.
Function calling Connect to external tools, APIs, and systems.
Structured output Return responses in structured formats like JSON.
Web search Search the web for up-to-date information.
File upload Process uploaded documents, spreadsheets, and other files.

Common questions

What is MiniMax-M3 good for?

Long multimodal work with text, images and video, which MiniMax says it was trained on from the start. Open weights under the MiniMax Community License, on a multi-GPU node.

When is MiniMax-M3 not a good fit?

Self-hosting on modest hardware: it takes a far bigger multi-GPU node than M2.7. No maximum output is published, and the MiniMax Community License has its own terms rather than Apache 2.0 or MIT.

What is the cheapest way to run MiniMax-M3?

Hosted, unless you push serious volume. MiniMax charges $0.30 in / $1.20 out per 1M tokens. The cheapest rental that fits is 8x RTX A6000 at $3,398 a month, which costs the same as about 8B tokens a month on that API.

How does MiniMax-M3 compare with MiniMax-M3.1-Flash?

MiniMax says M3.1-Flash keeps MiniMax-M3's mixture-of-experts design but swaps its first three full-attention layers for sparse attention, adds quantization-aware training and a new speculative decoding head, and does much better on agent, coding, computer use and security tasks. Same 1M-token context and text, image and video input. MiniMax-M3 lists at 100% more per 1M input tokens and 100% more per 1M output tokens than MiniMax-M3.1-Flash. It came out 4 months earlier.

Can I self-host MiniMax-M3?

Yes. The weights are released under the MiniMax Community License. At 4-bit it needs about 241 GB of GPU memory, which starts at roughly $3,398 a month on the cheapest rental that fits.