What GPU do I need to run MiniMaxAI/MiniMax-M2?
A 230B (MoE) model built for coding and agentic tool-use workflows. 228.7B parameters, published in F8_E4M3. View on Hugging Face
MiniMax-M2 is published by MiniMaxAI on Hugging Face, with 258,299 downloads and 1,503 likes to date. It's a MiniMaxM2ForCausalLM model built for text-generation, published natively in F8_E4M3.
What MiniMax-M2 is
MiniMax-M2 is a 230B-parameter mixture-of-experts language model published by MiniMax AI on Hugging Face, built for coding and agentic workflows with 10B active parameters per token. It is released under a modified MIT license.
License note: based on the MIT license with MiniMax's own modifications; read the linked LICENSE before commercial use. Facts in this section are sourced from MiniMax-M2's Hugging Face model card, not benchmarked by Aquanode.
What it's used for
- Coding agents and multi-file edits
- Agentic tool use and long-horizon tasks
- Chat assistants
VRAM required & cheapest live GPU fit
Required VRAM = weight size at each precision times a flat 1.2 overhead for activations and fragmentation. The KV-cache grows with context and is not in that factor; it is listed per model below where the architecture is published. Full formula and assumptions: methodology.
| Precision | Weight size | Required VRAM | Cheapest live fit | GPUs needed | Est. $/hr (full fit) |
|---|---|---|---|---|---|
| FP8 (native) | 213.0 GB | 255.6 GB | RTX 5000 Ada | 8 | $4.31/hr |
| INT4 (quantized) | 106.5 GB | 127.8 GB | AMD MI300X | 1 | $2.63/hr |
| cheaper alt. | RTX A5000 | 6 | $1.06/hr |
A GPU is only matched to a row if its hardware supports that precision, and the primary recommendation is always a single-GPU fit when one exists.
INT4 requires a quantized checkpoint actually published for this model, check its Hugging Face page before relying on this row.
Cheapest way to run MiniMax-M2 at its published (F8_E4M3) precision: 8× RTX 5000 Ada, at $0.539/hr per GPU ($4.31/hr total). Quantizing to FP8 or INT4 (rows above) can cost less, but requires a compatible quantized checkpoint to exist for this model.
MiniMax-M2: KV cache by context length
The KV cache is the memory the attention layers hold for every token of context, on top of the weights. It is not part of the flat 1.2x overhead in the table above, grows with context length and with every concurrent request, and is why a long-context deployment needs more VRAM than the table shows.
Grouped-query attention: every layer caches keys and values for a few shared KV heads. Cached per token: 253,952 bytes at 16-bit.
| Context | KV cache, one sequence |
|---|---|
| 4K | 0.97 GB |
| 32K | 7.8 GB |
| 128K | 31.0 GB |
| 192K (model maximum) | 46.5 GB |
Computed from the layer, head and window counts in the model's own configuration (MiniMaxAI/MiniMax-M2), at 16 bits per cached value. Each concurrent sequence needs its own cache, on top of the weights.
MiniMax-M2: common questions
Can MiniMax-M2 run on a single GPU?
No. At FP8 (native) it needs 255.6 GB of VRAM, more than a single 32 GB desktop card holds. The cheapest capable card in the live feed is a 32.0 GB RTX 5000 Ada, and it takes 8 of them.
Is MiniMax-M2 already quantized?
Yes. It is published in FP8, one byte per parameter, so the 255.6 GB figure is already a quantized footprint rather than a full-precision one. Only INT4 goes below it, at 127.8 GB. A GPU without FP8 tensor cores cannot run it as published, which is why cards here are matched on precision support and not on VRAM alone.
How many GPUs do I need to run MiniMax-M2?
8 at FP8 (native). It needs 255.6 GB of VRAM and the cheapest capable live offer is a 32.0 GB RTX 5000 Ada, so 8 of them come to $4.31/hr in total.
Does quantizing MiniMax-M2 lower the GPU bill?
Yes. At FP8 (native) the cheapest live fit is 8 RTX 5000 Ada cards at $4.31/hr. At INT4 (quantized) it drops to one AMD MI300X at $2.63/hr, provided a quantized checkpoint exists for it.
How to run MiniMax-M2
Run MiniMax-M2 with vLLM
From MiniMaxAI/MiniMax-M2's own vLLM deployment guide (8-GPU recipe).
SAFETENSORS_FAST_GPU=1 vllm serve \
MiniMaxAI/MiniMax-M2 --trust-remote-code \
--enable_expert_parallel --tensor-parallel-size 8 \
--enable-auto-tool-choice --tool-call-parser minimax_m2 \
--reasoning-parser minimax_m2_append_thinkSource: https://huggingface.co/MiniMaxAI/MiniMax-M2/raw/main/docs/vllm_deploy_guide.md
Deploy MiniMax-M2 on Aquanode
Aquanode has no one-click deploy template for MiniMax-M2; you install the inference engine yourself with the commands below. Aquanode sells GPU pods billed per second, not a hosted inference API.
- Launch a bare GPU pod sized to the requirement above (8× RTX 5000 Ada or larger).
- Open a terminal on the pod, or save one of the commands above as a startup script so it runs automatically the first time the pod boots.
- Run the command and connect to the resulting endpoint.
Launch a GPU podWeight-to-VRAM math, the fit rules, and how live prices are normalized: full methodology.
More MiniMax M2 models
- MiniMax-M2.7 (228.7B, F8_E4M3)
- MiniMax-M2.1 (228.7B, F8_E4M3)
- MiniMax-M2.5 (228.7B, F8_E4M3)
Alternatives at this size
Other models for text-generation within about a third of MiniMax-M2's 228.7B parameters, from other model lines.
- MiMo-V2.5 (310.8B, F8_E4M3)
- Qwen3-235B-A22B (235.1B, BF16)
- Step-3.5-Flash (199.4B, BF16)
- Hy3-preview (298.8B, BF16)
- DeepSeek-V2 (235.7B, BF16)
More on MiniMax-M2
Fits on a 288 GB GPU at FP8: every model that fits in 288 GB.
Fits on a 141 GB GPU at INT4: every model that fits in 141 GB.
Best chat and assistants models: how MiniMax-M2 ranks against the rest.
Compare MiniMax-M2 with Qwen3-235B-A22B, GLM-4.5, Qwen3-235B-A22B-Instruct-2507, GLM-4.7, Hy3-preview.
Related reading: H100 pricing and specs, How much VRAM you need for LLMs, Serving LLMs with vLLM, vLLM vs TensorRT-LLM vs SGLang, and Best GPU for LLM inference.
