MiniMax M3 — Hardware Requirements & GPU Compatibility
VisionFunctions
MiniMax-M3 is MiniMax's native multimodal model, trained on text, image, and video together from the very first training step rather than adding vision later. It has about 427 billion total parameters with roughly 23 billion activated per token, and uses a custom sparse-attention mechanism to keep long-context inference efficient, alongside three selectable reasoning modes for trading latency against answer quality on coding and agentic "cowork" tasks. It supports a one-million-token context window and is released under a custom license. At this scale, 4-bit quantization needs roughly 246GB of memory, so local use requires server-class or multi-GPU hardware; most users will reach it through a hosted endpoint.
MiniMax M3 runs on a 512 GB+ Mac or a datacenter GPU
Best local pick → Mac Studio (M3 Ultra, 512GB) · Q4_K_M · 258 GB · ~31 tok/s
Macs from 128 GB (IQ2_XXS · ~46 tok/s)
Score 112.5#21 of 100 open#18 of 96 that fit ≤512 GBOther
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
llmrun Composite Score →
112.5
- 90% range 109.7–115.3
- #21 of 100 open models
- #18 of 96 that fit ≤512 GB
Run MiniMax M3
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
ollama run hf.co/unsloth/MiniMax-M3-GGUF:Q4_K_MQ4_K_M runs on one Instinct MI350X 288 GB · 16K context
Download: UD-Q4_K_M/MiniMax-M3-UD-Q4_K_M-00001-of-00007.gguf264.93 GB + 6 partsunsloth
Other ways
each opens step-by-stepRun in cloud
Fits on B300 (288 GB) (30 GB headroom) · Q4_K_M
- Generation speed
- ~179 tok/s
- generation speed
- Cost per 1M output tokens
- $10.78
- per 1M output tokens
How Much VRAM Does MiniMax M3 Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | Runs on | |
|---|---|---|---|---|
| 182.0 GB | 184.8 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Runs on B200 192 GB | |
| 187.3 GB | 190.2 GB | 3-bit small — 3-bit small quantization | Runs on B200 192 GB | |
| 208.7 GB | 211.5 GB | 3-bit medium — 3-bit medium quantization | Runs on Instinct MI325X 256 GB | |
| 214.0 GB | 216.8 GB | 4-bit · legacy — 4-bit legacy quantization | Runs on Instinct MI325X 256 GB | |
| 256.7 GB | 259.5 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Runs on Instinct MI350X 288 GB | |
| 304.8 GB | 307.6 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Runs on 5× H100 SXM 80 GB | |
| 352.8 GB | 355.6 GB | 6-bit, very good — 6-bit quantization, very good quality | Runs on 5× H100 SXM 80 GB | |
| 427.5 GB | 430.4 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Runs on 6× H100 SXM 80 GB |
Selected Q4_K_M
4-bit medium quantization258.0GBat 16K context
- Weights
- 256.2 GB
- KV cache
- 1.5 GB
- Overhead
- 0.3 GB
Server-class: ~258 GB at 16K context · one Instinct MI350X 288 GB
Which Hardware Can Run MiniMax M3?
Q4_K_M · 256.7 GBMiniMax M3 (Q4_K_M) requires 256.7 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 334+ GB is recommended. Using the full 1049K context window can add up to 96.5 GB, bringing total usage to 353.2 GB. No consumer GPU has enough memory. 1 device with unified memory can run MiniMax M3, including Mac Studio (M3 Ultra, 512GB).
128–144 GB
Q4_K_M · best fit IQ2_XXS (119.2 GB)
4 GPUs · 2 Mac classes · 1 device family
GH200 Grace Hopper Superchip~199Simulate output on GH200 Grace Hopper SuperchipH200 NVL~198Simulate output on H200 NVLH200 SXM~198Simulate output on H200 SXMM2 Ultra*128 GB~46Simulate output on Mac Pro M2 Ultra (128 GB)M5 Max128 GB~40Simulate output on Mac Studio M5 Max (128 GB)GB10~29Simulate output on NVIDIA DGX Spark192+ GB
Q4_K_M · best fit Q3_K_S (188.6 GB)
6 GPUs · 1 Mac class · 2 device families
B200~201Simulate output on B200B300~201Simulate output on B300Instinct MI350X~72Simulate output on Instinct MI350XM3 Ultra*256–512 GB~37Simulate output on Mac Studio (M3 Ultra, 256GB)8x H100 SXM5~263Simulate output on NVIDIA DGX H1008x A100 SXM4~243Simulate 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 ≥ 119.2 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →
How good is MiniMax M3?
The llmrun Score merges 22 public benchmark boards into a single figure, where 100 represents the average for open models.
Score vs VRAM in the ≤512 GB tier
- MiniMax M3
- Other models
- Best Score for its VRAM
#18 of 96 that fit ≤512 GB
Each dot is a model at the best quant that fits ≤512 GB. The line joins the models with the best Score for their VRAM; a dot below it is beaten by a model that needs no more memory.
Full ≤512 GB ranking →Benchmarks
The 6 boards that weigh most in the llmrun Score, of 28 with a result for this model; badges show who ran the test.
- GPQA Diamond
Epoch AI · Jun 2026Independent
Independent90.9%
#6/98 open
- LMArena Text
LMArena · Oct 2026Arena
Arena1441.3
#19/187 open
- LiveBench Coding
LiveBench · Jun 2026Independent
Independent68.2
#19/19 open
- LiveBench Reasoning
LiveBench · Jun 2026Independent
Independent74.5
#17/19 open
- AIME 2024/2025
Epoch AI · Jun 2026Independent
Independent71.1%
#27/86 open
- LMArena WebDev
LMArena · Oct 2026Arena
Arena1482.5
#17/52 open
In the ≤512 GB class
96 models fit ≤512 GBSame Score, only models that fit ≤512 GB at their best quant: this model's peers by memory need.
- #1
MiMo V2.6 Pro RLXiaomi
133.4
499.3 GB
- #2
MiMo V2.6 Flash RLXiaomi
131.9
186.5 GB
- #3
GLM 5.3Z.ai
130.1
453.8 GB
- #4
DeepSeek V4.1 FlashDeepSeek
129.5
458.4 GB
- #5
DeepSeek V4 Flash 0731DeepSeek
127.6
183.0 GB
- #6
125.3
461.5 GB
Alternatives to MiniMax M3
MiniMax family · 9 models →Official models with an llmrun Score, each answering one question; sizes are each model's default quant (usually Q4_K_M) with a 16K-token context.
Best alternative at this size
GLM 5.3 Flash
Z.ai
- 11.7 higher Score
- 52.9 GB less VRAM
Score 124.2 · 205.2 GB
Other vendor
MiMo V2.6 Flash RL
Xiaomi
- 19.4 higher Score
- 71.5 GB less VRAM
Score 131.9 · 186.5 GB
Similar models
10| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| GLM 5.3Z.ai | 130.1 | 453.8 GB | —needs > 24 GB | similar score |
| DeepSeek V4.1 FlashDeepSeek | 129.5 | 458.4 GB | —needs > 24 GB | similar score |
| DeepSeek V4 Flash 0731DeepSeek | 127.6 | 183 GB | —needs > 24 GB | similar score |
| GLM 5.2Z.ai | 122.9 | 453.8 GB | —needs > 24 GB | similar score |
| Inkling SmallThinking Machines | 119.6 | 162.7 GB | —needs > 24 GB | similar score |
| GLM 5.1Z.ai | 117.8 | 454.1 GB | —needs > 24 GB | similar score |
| GLM 5Z.ai | 111.6 | 454.1 GB | —needs > 24 GB | similar score |
| Qwen3.6 27BAlibaba | 111.2 | 17.9 GB | ~38 | similar score |
| DeepSeek V3.2DeepSeek | 110.4 | 412.7 GB | —needs > 24 GB | similar score |
| NVIDIA Nemotron 3 Ultra 550B A55B BF16NVIDIA | 109.9 | 378.4 GB | —needs > 24 GB | similar score |
Fine-tunes & variants(1)
Specifications
- Publisher
- MiniMax
- Family
- MiniMax
- Parameters
- 427.0B
- Active parameters
- 19.8B per token
- Architecture
- MiniMaxM3SparseForConditionalGeneration
- Context Length
- 1,048,576 tokens
- Vocabulary Size
- 200,064
- Release Date
- 2026-06-02
- License
- Other
Frequently Asked Questions
How much VRAM does MiniMax M3 need?
MiniMax M3 requires 256.7 GB of VRAM at Q4_K_M, or 854.6 GB at BF16. Full 1049K context adds up to 96.5 GB (353.2 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 427.0B × 4.8 bits ÷ 8 = 256.2 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 ≈ 97 GB (at full 1049K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run MiniMax M3?
No — MiniMax M3 requires at least 117.9 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for MiniMax M3?
For MiniMax M3, Q4_K_M (256.7 GB) offers the best balance of quality and VRAM usage. Q5_K_S (294.1 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 117.9 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run MiniMax M3 on a Mac?
Yes — Mac Pro M2 Ultra (128 GB) and 14 other Macs can run MiniMax M3, 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 256.7 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 512GB).
Can I run MiniMax M3 locally?
Yes — MiniMax M3 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is MiniMax M3?
At Q4_K_M, MiniMax M3 can reach ~69 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 → 11.9 GB active ÷ (8000 × 0.65) = 2.29 ms, plus 60 layers × 0.055 ms = 3.30 ms, so 1000 ÷ 5.59 ms = ~179 tok/s
Estimated speed at Q4_K_M (256.7 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?
At Q4_K_M, the download is about 256.22 GB. The full-precision BF16 version is 854.08 GB. The smallest option (IQ2_XXS) is 117.44 GB.
Which GPUs can run MiniMax M3?
No single consumer GPU has enough VRAM to run MiniMax M3 at Q4_K_M (256.7 GB). Multi-GPU or professional hardware is required.
Which devices can run MiniMax M3?
1 device with unified memory can run MiniMax M3 at Q4_K_M (256.7 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.