Can I run MiniMax M2.5 locally?
MiniMax M2.5 by MiniMaxAI has 229B parameters (11B active per token, 256 experts, 8 used per token). At 4-bit with an 8K context it needs about 132 GB of GPU memory. It fits on 9 of 65 GPUs and Macs we track.
Memory needed by quant and context length
GB of GPU memory = weights + KV cache (FP16) + runtime overhead. Weights use real file sizes from unsloth/MiniMax-M2.5-GGUF where available (marked •), otherwise parameter count × bits per weight. How we calculate.
| Quant | Quality | 4K ctx | 8K ctx | 32K ctx | 128K ctx |
|---|---|---|---|---|---|
| 2-bit (Q2_K) • | rough | 79.3 GB | 80.3 GB | 86.1 GB | 109 GB |
| 3-bit (Q3_K_M) • | usable | 104 GB | 104 GB | 110 GB | 134 GB |
| 4-bit (Q4_K_M) • | good, most popular | 131 GB | 132 GB | 137 GB | 161 GB |
| 5-bit (Q5_K_M) • | very good | 153 GB | 154 GB | 160 GB | 183 GB |
| 6-bit (Q6_K) • | near lossless | 177 GB | 178 GB | 183 GB | 207 GB |
| 8-bit (Q8_0) • | effectively lossless | 228 GB | 229 GB | 235 GB | 258 GB |
| 16-bit (F16/BF16) • | original weights | 428 GB | 429 GB | 435 GB | 458 GB |
Which GPUs and Macs run MiniMax M2.5?
Best quality quant that fits at 8K context. "Tight" = within 10% of usable memory. Speed is a ceiling from memory bandwidth (real speed is typically 50–80% of it).
Prefer an API?
MiniMax: MiniMax M2.5 is on OpenRouter at $0.295 per 1M input tokens and $1.20 per 1M output tokens. Price details and history →
Questions
How much VRAM does MiniMax M2.5 need?
About 132 GB at 4-bit (Q4_K_M) with an 8K context: 129 GB of weights (real GGUF file size), 1.9 GB of KV cache and 0.8 GB of runtime overhead. At 8-bit it needs about 229 GB.
What is the smallest GPU that runs MiniMax M2.5?
No single consumer GPU in our table fits it at Q4_K_M with 8K context; the smallest option that does is the Mac M3 Ultra 256GB (256 GB).
Can I run MiniMax M2.5 on a Mac?
Yes. At Q4_K_M with 8K context it fits on a Mac M3 Ultra 256GB or larger, using the default macOS GPU memory limit.
Similar models
Same family or a similar parameter count, with the VRAM each needs at Q4_K_M and 8K context.
MiniMax M2.7
229B · 133 GB at 4-bitMiniMax M2
229B · 132 GB at 4-bitMiniMax M2.1
229B · 132 GB at 4-bitDeepSeek V2 Chat
236B · 135 GB at 4-bitIntern S1
241B · 139 GB at 4-bitStep 3.7 Flash
201B · 119 GB at 4-bitModel facts
| Hugging Face repo | MiniMaxAI/MiniMax-M2.5 |
| Parameters | 228,689,764,864 (229B) · ~11B active (computed from config) |
| Layers / hidden size | 62 / 3072 |
| Attention | GQA; 48 heads, 8 KV heads, head dim 128 |
| KV cache per token (FP16) | 248.0 KiB |
| Max context | 196,608 tokens |
| License | other |
| Released on HF | Feb 12, 2026 |
| Downloads (30d) / likes | 1,027,114 / 1,532 |
Sources: Hugging Face model API · config.json (MiniMaxAI/MiniMax-M2.5) · GGUF files · GPU specs from vendor pages (list). Data fetched Nov 17, 2026. Spot an error? Report it.