MiniMax M3.1 Flash has 427.1B parameters (23B active per token). In BF16 its weights alone take 795 GiB; quantized to 4-bit, about 240 GiB. Each 32K-token request adds 4.22 GiB of KV cache. Cheapest way to run it today: 6 × RTX A6000 (48 GB) at $2.00 per hour. It reads text, images and video, calls tools and has a reasoning mode.
Weights plus the KV cache of one request (BF16 cache), in GiB. Add 1 to 3 GiB for the inference engine itself.
Weights format
Weights only
+ 8K context
+ 32K context
+ 128K context
+ 1M context
NVFP4 experts + MXFP8 (native)
233
234
237
250
368
BF16 / FP16
795
796
800
812
930
FP8 / INT8
400
401
404
417
535
4-bit (GGUF Q4_K_M)
240
241
244
257
375
How many GPUs to run MiniMax M3.1 Flash
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
GPU
NVFP4 experts + MXFP8 (native)
BF16 / FP16
8-bit
4-bit
GeForce RTX 4090 24 GB
Not on this GPU
48 (several machines)
24 (several machines)
16 (several machines)
GeForce RTX 5090 32 GB
Not on this GPU
32 (several machines)
16 (several machines)
16 (several machines)
RTX PRO 6000 Blackwell 96 GB
Not on this GPU
16 (several machines)
5
3
Mac M5 Max 128 GB 96 GB usable
Not on this GPU
16 (several machines)
5
3
Ryzen AI Max+ 395 128 GB 96 GB usable
Not on this GPU
16 (several machines)
5
3
DGX Spark (GB10) 128 GB
Not on this GPU
7
4
2
NVIDIA H100 80 GB
Not on this GPU
16 (several servers)
7
4
NVIDIA H200 141 GB
Not on this GPU
7
4
2
NVIDIA B200 (HGX) 180 GB
2
6
3
2
8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size. Not on this GPU: the official NVFP4 experts + MXFP8 weights do not run there (llama.cpp needs a GGUF version, and vLLM needs a GPU with FP4 support).
Cheapest way to run MiniMax M3.1 Flash today
For each GPU: the fewest cards that fit MiniMax M3.1 Flash with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-23).
On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
Reasoning. Has a reasoning mode. Reasoning tokens take context and KV cache like any other token: leave room for them.
KV cache: how memory grows with context
Architecture: MoE 128 experts + 1 shared · 60 layers, MiniMax Sparse Attention (GQA 4 KV × 128, 128-dim index keys) · vision encoder. Its cache stays compact because MLA stores one compressed latent per token and layer instead of full keys and values. Only 23B parameters are active per token, but all 427B must sit in memory: VRAM depends on the total, speed on the active part.
Context per request
BF16 cache
FP8 cache
8K tokens
1.05 GiB
0.59 GiB
32K tokens
4.22 GiB
2.34 GiB
128K tokens
16.9 GiB
9.38 GiB
1M tokens
135 GiB
75.0 GiB
MiniMax M3.1 Flash VRAM FAQ
How much VRAM does MiniMax M3.1 Flash need?
In BF16 the weights alone take 795 GiB (854 GB). In FP8 that is 400 GiB, and about 240 GiB with 4-bit quantization (Q4_K_M); the official NVFP4 experts + MXFP8 checkpoint is 233 GiB. Each request then adds KV cache: 1.05 GiB at 8K tokens and 4.22 GiB at 32K (BF16 cache).
Can MiniMax M3.1 Flash run on a single RTX 4090 (24 GB)?
No. Even in 4-bit it needs 16 RTX 4090s, spread over several machines. In 4-bit it needs 4 H100 GPUs.
Can MiniMax M3.1 Flash run on an RTX 5090 (32 GB)?
No. Even in 4-bit it needs 16 RTX 5090s, spread over several machines.
Can MiniMax M3.1 Flash run on a Mac?
Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 512 GB of unified memory, of which macOS gives the GPU about 384 GB, for one request with an 8K-token context. In 8-bit it does not fit on a single Mac.
Can MiniMax M3.1 Flash run on a DGX Spark or a Ryzen AI Max+ 395 (128 GB)?
No. Even in 4-bit it needs 2 DGX Spark units or 3 Ryzen AI Max+ 395 machines (128 GB), linked over the network.
What is the cheapest way to run MiniMax M3.1 Flash?
On 2026-10-23, the cheapest on-demand setup is 6 × RTX A6000 (48 GB) with Q4_K_M weights on Vast.ai, at $2.00 per hour (about $1,462.92 per month). Next: 7 × NVIDIA L40S (48 GB) with 4-bit weights on Vast.ai, at $3.27 per hour. Sized for one 8K-token request; prices are checked every hour.
How many H100 GPUs does MiniMax M3.1 Flash need?
With one request and an 8K-token context: 16 in BF16, 7 in FP8 and 4 in 4-bit (vLLM, 90% of memory usable). Above 8 GPUs, that means several servers. Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
Can MiniMax M3.1 Flash read images?
Yes. MiniMax M3.1 Flash reads text, images and video. Images add tokens to the context, so they also add KV cache.
Does MiniMax M3.1 Flash support tool calling?
Yes, MiniMax M3.1 Flash supports tool (function) calling: it can ask your application to run a function and use the result.
Does MiniMax M3.1 Flash have a thinking mode?
Has a reasoning mode. Reasoning tokens count toward the context and the KV cache.
MiniMax M3.1 Flash VRAM badge
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/minimax-m3-1-flash.json.
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.
MiniMax, China · Released October 22, 2026