Mixtral 34Bx2 MoE 60B — Hardware Requirements & GPU Compatibility
Chat
8.2K downloads
113 likesSpecifications
- Publisher
- cloudyu
- Family
- Mixtral
- Parameters
- 60.8B
- Architecture
- MixtralForCausalLM
- Context Length
- 200,000 tokens
- Vocabulary Size
- 64,000
- Release Date
- 2026-01-06
- License
- Apache 2.0
Get Started
HuggingFace
How Much VRAM Does Mixtral 34Bx2 MoE 60B Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 122.4 GB | 171.1 GB | 121.63 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Mixtral 34Bx2 MoE 60B?
BF16 · 122.4 GBMixtral 34Bx2 MoE 60B (BF16) requires 122.4 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 160+ GB is recommended. Using the full 200K context window can add up to 48.7 GB, bringing total usage to 171.1 GB. No single GPU has enough memory — multi-GPU or cluster setups are needed.
Which Devices Can Run Mixtral 34Bx2 MoE 60B?
BF16 · 122.4 GB5 devices with unified memory can run Mixtral 34Bx2 MoE 60B, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, Mac Studio M4 Max (128 GB).
Runs great
— Plenty of headroom
Decent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does Mixtral 34Bx2 MoE 60B need?
Mixtral 34Bx2 MoE 60B requires 122.4 GB of VRAM at BF16. Full 200K context adds up to 48.7 GB (171.1 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 60.8B × 16 bits ÷ 8 = 121.6 GB
KV Cache + Overhead ≈ 0.8 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 49.5 GB (at full 200K context)
VRAM usage by quantization
BF16122.4 GBBF16 + full context171.1 GB- Can NVIDIA GeForce RTX 5090 run Mixtral 34Bx2 MoE 60B?
No — Mixtral 34Bx2 MoE 60B requires at least 122.4 GB at BF16, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
- Can I run Mixtral 34Bx2 MoE 60B on a Mac?
Mixtral 34Bx2 MoE 60B requires at least 122.4 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Mixtral 34Bx2 MoE 60B locally?
Yes — Mixtral 34Bx2 MoE 60B can run locally on consumer hardware. At BF16 quantization it needs 122.4 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Mixtral 34Bx2 MoE 60B?
At BF16, Mixtral 34Bx2 MoE 60B can reach ~24 tok/s on AMD Instinct MI300X. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 122.4 × 0.55 = ~24 tok/s
Estimated speed at BF16 (122.4 GB)
AMD Instinct MI300X~24 tok/sAMD Instinct MI250X~15 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Mixtral 34Bx2 MoE 60B?
At BF16, the download is about 121.63 GB.
- Which GPUs can run Mixtral 34Bx2 MoE 60B?
No single consumer GPU has enough VRAM to run Mixtral 34Bx2 MoE 60B at BF16 (122.4 GB). Multi-GPU or professional hardware is required.
- Which devices can run Mixtral 34Bx2 MoE 60B?
5 devices with unified memory can run Mixtral 34Bx2 MoE 60B at BF16 (122.4 GB), including Mac Pro M2 Ultra (192 GB), Mac Studio M2 Ultra (192 GB), Mac Studio M4 Max (128 GB), NVIDIA DGX A100 640GB. Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.