Models / MiniMax M3

MiniMax M3

premier
Find for my hardware
coding agentic vision multimodal long-context
Context
1048k
License
minimax community
Developer
MiniMax
Released
Jun 2026

Related models

MiniMax M2.7
workstation 61GB min RAM Apr 9, 2026

Save your hardware and every model page answers the real question: will it run on your machine, and how fast?

Join free - save your rig →

Run it locally

Per-quant memory needs and a static "can you run it?" reference - no rig entry required

UD-IQ1_M
128.4GB 128.4GB min 152.0GB rec
Quantized build
UD-Q2_K_XL
143.0GB 143.0GB min 168.0GB rec
Unsloth Dynamic 2-bit XL - best quality-per-GB in the Dynamic line
MXFP4
256.5GB 256.5GB min 282.0GB rec
Quantized build
UD-Q4_K_XL
264.9GB 264.9GB min 290.0GB rec
Unsloth Dynamic 4-bit XL - lossless at Q4 size

The reference hardware

Schematic of the NVIDIA Jetson Orin NX 16GB reference rig - 16GB unified memory, 102 GB/s aggregate bandwidth
Schematic of the Jetson AGX Orin 64GB reference rig - 64GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Single GTX 1080 Ti (11GB) reference rig - 11GB VRAM, 484 GB/s aggregate bandwidth
Schematic of the Single RTX 4090 (24GB) reference rig - 24GB VRAM, 1008 GB/s aggregate bandwidth
Schematic of the Single RTX 5090 (32GB) reference rig - 32GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the RTX PRO 6000 Blackwell (96GB) reference rig - 96GB VRAM, 1792 GB/s aggregate bandwidth
Schematic of the 2x RTX 3090 (48GB) reference rig - 48GB VRAM, 1872 GB/s aggregate bandwidth
Schematic of the 2x RTX 5090 (64GB) reference rig - 64GB VRAM, 3584 GB/s aggregate bandwidth
Schematic of the 4x RTX 4090 (96GB) reference rig - 96GB VRAM, 4032 GB/s aggregate bandwidth
Schematic of the 4x RTX 5090 (128GB) reference rig - 128GB VRAM, 7168 GB/s aggregate bandwidth
Schematic of the MacBook Pro M5 Max 128GB reference rig - 128GB unified memory, 614 GB/s aggregate bandwidth
Schematic of the Ryzen AI Max+ 395 128GB reference rig - 128GB unified memory, 256 GB/s aggregate bandwidth
Schematic of the DGX Spark 128GB unified reference rig - 128GB unified memory, 273 GB/s aggregate bandwidth
Schematic of the 4x H100 80GB (320GB) reference rig - 320GB VRAM, 13400 GB/s aggregate bandwidth
Schematic of the NVIDIA DGX Station 748GB reference rig - 748GB unified memory, 8000 GB/s aggregate bandwidth
Schematic of the 8x RTX 3090 rack (192GB) reference rig - 192GB VRAM, 7489 GB/s aggregate bandwidth
Schematic of the AMD Instinct MI300X (192GB) reference rig - 192GB VRAM, 5324 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 192GB reference rig - 192GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Mac Studio M4 Ultra 512GB reference rig - 512GB unified memory, 1092 GB/s aggregate bandwidth
Schematic of the Dual EPYC 9004 + 768GB DDR5-4800 reference rig - 768GB unified memory, 460 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-3200 + 2x RTX 3090 reference rig - 560GB unified memory, 204 GB/s aggregate bandwidth
Schematic of the Epyc + 512GB DDR4-2400 + 2x RTX 3090 reference rig - 560GB unified memory, 153 GB/s aggregate bandwidth

22 reference configs, drawn in-house. Scroll for more.

Can you run it? - reference rigs

Rig UD-IQ1_M UD-Q2_K_XL MXFP4 UD-Q4_K_XL
NVIDIA Jetson Orin NX 16GB no -> cloud no -> cloud no -> cloud no -> cloud
Jetson AGX Orin 64GB no -> cloud no -> cloud no -> cloud no -> cloud
Single GTX 1080 Ti (11GB) no -> cloud no -> cloud no -> cloud no -> cloud
Single RTX 4090 (24GB) no -> cloud no -> cloud no -> cloud no -> cloud
Single RTX 5090 (32GB) no -> cloud no -> cloud no -> cloud no -> cloud
RTX PRO 6000 Blackwell (96GB) offload offload no -> cloud no -> cloud
2x RTX 3090 (48GB) no -> cloud no -> cloud no -> cloud no -> cloud
2x RTX 5090 (64GB) no -> cloud no -> cloud no -> cloud no -> cloud
4x RTX 4090 (96GB) offload offload no -> cloud no -> cloud
4x RTX 5090 (128GB) offload offload no -> cloud no -> cloud
MacBook Pro M5 Max 128GB no -> cloud no -> cloud no -> cloud no -> cloud
Ryzen AI Max+ 395 128GB no -> cloud no -> cloud no -> cloud no -> cloud
DGX Spark 128GB unified no -> cloud no -> cloud no -> cloud no -> cloud
4x H100 80GB (320GB) fast 57.2t/s fast 51.4t/s fast 28.7t/s fast 27.8t/s
NVIDIA DGX Station 748GB fast 34.1t/s fast 30.7t/s ok 17.1t/s ok 16.6t/s
8x RTX 3090 rack (192GB) fast 32.0t/s fast 28.7t/s offload offload
AMD Instinct MI300X (192GB) fast 22.7t/s fast 20.4t/s offload offload
Mac Studio M4 Ultra 192GB slow 5.1t/s slow 4.6t/s no -> cloud no -> cloud
Mac Studio M4 Ultra 512GB slow 5.1t/s slow 4.6t/s slow 2.6t/s slow 2.5t/s
Dual EPYC 9004 + 768GB DDR5-4800 slow 2.0t/s slow 1.8t/s slow 1.0t/s slow 1.0t/s
Epyc + 512GB DDR4-3200 + 2x RTX 3090 slow 0.9t/s slow 0.8t/s slow 0.4t/s slow 0.4t/s
Epyc + 512GB DDR4-2400 + 2x RTX 3090 slow 0.7t/s slow 0.6t/s slow 0.3t/s slow 0.3t/s

Fit tiers use the same will-it-run logic as the rig finder. For comfortable fits, the badge reflects decode speed: fast >=20 t/s, ok 8-20 t/s, slow <8 t/s. t/s is a bandwidth estimate, not a measured benchmark.

Download options

UD-IQ1_M Unsloth - local-optimized
128.4GB dl 128.4GB min 152.0GB rec
REC RAM vs largest quant
128.4GB unsloth 1-bit weights (UD-IQ1_M, file sum of 4 shards) + KV cache; ~152GB to run
GGUF on HF →
UD-Q2_K_XL Unsloth - local-optimized
143.0GB dl 143.0GB min 168.0GB rec
REC RAM vs largest quant
143.0GB unsloth Dynamic 2-bit weights (UD-Q2_K_XL, file sum of 4 shards) + KV cache at 123MB per 1k tokens; ~168GB to run
GGUF on HF →
MXFP4 official
256.5GB dl 256.5GB min 282.0GB rec
REC RAM vs largest quant
256.5GB MXFP4 weights (unsloth MXFP4_MOE folder, file sum of 7 shards) + KV; ~282GB to run; served via ATOM on ROCm
GGUF on HF →
UD-Q4_K_XL Unsloth - local-optimized
264.9GB dl 264.9GB min 290.0GB rec
REC RAM vs largest quant
264.9GB unsloth Dynamic 4-bit weights (UD-Q4_K_XL, file sum of 7 shards) + KV; ~290GB to run
GGUF on HF →

Or run it in the cloud

No per-token API provider pricing tracked for MiniMax M3 yet. For flagship list prices, see the calculator.

PRICE HISTORY

Inference cost over time

Data accumulates from the first daily sync - longer ranges populate over time. Prices come from OpenRouter snapshots, not a historical API.

Loading price history...