MiniMax M3.1 Flash — Hardware Requirements & GPU Compatibility
VisionFunctions
MiniMax-M3.1-Flash is a native multimodal mixture-of-experts model from MiniMax, aimed at coding, agentic workflows, computer use, and cybersecurity. It accepts text, image, and video input and produces text. It has about 428 billion total parameters with 23 billion active per token, and builds on MiniMax-M3 with sparse attention throughout, quantization-aware training, and a speculative-decoding draft head. A Q4 build needs roughly 245 GB of memory, so it calls for a multi-GPU server or a very large unified-memory machine. The context window is up to 1 million tokens. It is released under the MiniMax Community license, and it was published on October 17, 2026. Its card reports 73.7% on DeepSWE v1.1, 61.7 on OSWorld V2, and 82.2 on CyberGym. Reasoning effort can be set from low to max in the chat template. The card recommends SGLang, vLLM, or TokenSpeed for serving.
MiniMax M3.1 Flash runs on a 512 GB+ Mac or a datacenter GPU
Best local pick → Mac Studio (M3 Ultra, 512GB) · Q4_K_M · 258.6 GB · ~28 tok/s
Macs from 128 GB (IQ2_XXS · ~43 tok/s)
Other
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run MiniMax M3.1 Flash
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/AMAImedia/MiniMax-M3.1-Flash-NVFP4-GGUF:Q4_K_MQ4_K_M runs on one Instinct MI350X 288 GB · 16K context
Download: GGUF/Q4_K_M/Sakura-MiniMax-M3.1-Flash-Q4_K_M-00001-of-00006.gguf258.10 GB + 5 partsAMAImedia
Other ways
each opens step-by-stepRun in cloud
Fits on B300 (288 GB) (29 GB headroom) · Q4_K_M
- Generation speed
- ~168 tok/s
- generation speed
- Cost per 1M output tokens
- $11.49
- per 1M output tokens
How Much VRAM Does MiniMax M3.1 Flash Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 118.2 GB | 121 GB | 2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss | Runs on Instinct MI250X 128 GB | |
| 128.9 GB | 131.7 GB | 2-bit · i-quant extra small — Importance-weighted 2-bit, extra small | Runs on H200 NVL 141 GB | |
| 182.4 GB | 185.2 GB | 3-bit · i-quant small — Importance-weighted 3-bit, small | Runs on B200 192 GB | |
| 182.4 GB | 185.2 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on B200 192 GB | |
| 209.1 GB | 212 GB | 3-bit medium — 3-bit medium quantization | Runs on Instinct MI325X 256 GB | |
| 257.3 GB | 260.1 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on Instinct MI350X 288 GB | |
| 305.4 GB | 308.3 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on 5× H100 SXM 80 GB | |
| 353.6 GB | 356.4 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on 5× H100 SXM 80 GB | |
| 428.5 GB | 431.3 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 6× H100 SXM 80 GB | |
| 856.5 GB | 859.3 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Runs on 7× H200 SXM 141 GB |
Selected Q4_K_M
4-bit medium quantization258.6GBat 16K context
- Weights
- 256.8 GB
- KV cache
- 1.5 GB
- Overhead
- 0.3 GB
Server-class: ~258.6 GB at 16K context · one Instinct MI350X 288 GB
Which Hardware Can Run MiniMax M3.1 Flash?
Q4_K_M · 257.3 GBMiniMax M3.1 Flash (Q4_K_M) requires 257.3 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 335+ GB is recommended. Using the full 1049K context window can add up to 96.4 GB, bringing total usage to 353.7 GB. No consumer GPU has enough memory. 1 device with unified memory can run MiniMax M3.1 Flash, including Mac Studio (M3 Ultra, 512GB).
128–144 GB
Q4_K_M · best fit IQ2_XXS (119.5 GB)
4 GPUs · 2 Mac classes · 1 device family
GH200 Grace Hopper Superchip~189Simulate output on GH200 Grace Hopper SuperchipH200 NVL~187Simulate output on H200 NVLH200 SXM~187Simulate output on H200 SXMM2 Ultra*128 GB~43Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~37Simulate output on Mac Studio M5 Max (128 GB)GB10~26Simulate output on NVIDIA DGX Spark192+ GB
Q4_K_M · best fit IQ3_S (183.7 GB)
6 GPUs · 1 Mac class · 2 device families
B200~193Simulate output on B200B300~193Simulate output on B300Instinct MI350X~71Simulate output on Instinct MI350XM3 Ultra*256–512 GB~34Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~259Simulate output on NVIDIA DGX H1008x A100 SXM4~237Simulate output on NVIDIA DGX A100 640GB
By VRAM tier · typical GPUs and Macs at Q4_K_M
Consumer GPUs (up to 32 GB) and Macs up to 96 GB: too small for this model (needs ≥ 119.5 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is MiniMax M3.1 Flash?
llmrun Score blends public benchmark boards into one number; 100 = the average open model.
Not enough data yet for an llmrun Score.
A Score needs results on enough independent boards. How the Composite Score is built
Benchmarks
The 6 boards that weigh most in the llmrun Score, of 8 with a result for this model; badges show who ran the test.
- LMArena WebDev
LMArena · Oct 2026Arena
Arena1588.8
#12/55 open
- LiveBench 2026 Coding
LiveBench · Oct 2026Independent
Independent83.9
#4/17 open
- LiveBench 2026 Long-Horizon
LiveBench · Oct 2026Independent
Independent34.7
#8/17 open
- LiveBench 2026 Design
LiveBench · Oct 2026Independent
Independent74.0
#8/17 open
- LiveBench 2026 Research
LiveBench · Oct 2026Independent
Independent72.7
#8/17 open
- LiveBench 2026 Browser
LiveBench · Oct 2026Independent
Independent92.0
#1/17 open
Alternatives to MiniMax M3.1 Flash
MiniMax family · 10 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| Qwen3.5 397B A17BAlibaba | 113.0 | 242.6 GB | —needs > 24 GB | same size |
| Llama 4 Maverick 17B 128E InstructMeta | 87.1 | 272.7 GB | —needs > 24 GB | same size |
| Llama 3.1 405B InstructMeta | 83.5 | 275.6 GB | —needs > 24 GB | same size |
| Intern S2 397BInternLM | — | 242.6 GB | —needs > 24 GB | same size |
| Ornith 1.5 397Bornith-ai | — | 242.6 GB | —needs > 24 GB | same size |
| Trinity Large ThinkingArcee AI | — | 799.6 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- MiniMax
- Family
- MiniMax
- Parameters
- 428B
- Active parameters
- 23B per token
- Architecture
- MiniMaxM3SparseForConditionalGeneration
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 200,064
- Release Date
- 2026-10-17
- License
- Other
Frequently Asked Questions
How much VRAM does MiniMax M3.1 Flash need?
MiniMax M3.1 Flash requires 257.3 GB of VRAM at Q4_K_M, or 856.5 GB at BF16. Full 1049K context adds up to 96.4 GB (353.7 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 428B × 4.8 bits ÷ 8 = 256.8 GB
KV Cache + Overhead ≈ 0.5 GB (at 2K context + ~0.3 GB framework)
Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.
KV Cache + Overhead ≈ 96.9 GB (at full 1049K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run MiniMax M3.1 Flash?
No — MiniMax M3.1 Flash requires at least 118.2 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for MiniMax M3.1 Flash?
For MiniMax M3.1 Flash, Q4_K_M (257.3 GB) offers the best balance of quality and VRAM usage. Q5_K_M (305.4 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 118.2 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run MiniMax M3.1 Flash on a Mac?
Yes — Mac Pro M2 Ultra (128 GB) and 14 other Macs can run MiniMax M3.1 Flash, the smallest at IQ2_XXS. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 257.3 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 512GB).
Can I run MiniMax M3.1 Flash locally?
Yes — MiniMax M3.1 Flash can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is MiniMax M3.1 Flash?
At Q4_K_M, MiniMax M3.1 Flash can reach ~67 tok/s on AMD Instinct MI350X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = 1000 ÷ (active GB ÷ (bandwidth GB/s × efficiency) × 1000 + layers × routing ms)
Mixture-of-Experts: just the active experts get read for each token, along with a fixed routing cost per layer.
Example: NVIDIA B300 → 13.8 GB active ÷ (8000 × 0.65) = 2.66 ms, plus 60 layers × 0.055 ms = 3.30 ms, so 1000 ÷ 5.96 ms = ~168 tok/s
Estimated speed at Q4_K_M (257.3 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of MiniMax M3.1 Flash?
At Q4_K_M, the download is about 256.80 GB. The full-precision BF16 version is 856.00 GB. The smallest option (IQ2_XXS) is 117.70 GB.
Which GPUs can run MiniMax M3.1 Flash?
No single consumer GPU has enough VRAM to run MiniMax M3.1 Flash at Q4_K_M (257.3 GB). Multi-GPU or professional hardware is required.
Which devices can run MiniMax M3.1 Flash?
1 device with unified memory can run MiniMax M3.1 Flash at Q4_K_M (257.3 GB), including Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.