V LLMVRAMCalc
EN中文ESDE
Calculator › Qwen3.5 122B-A10B › Mac Studio M4 Ultra 256GB

Qwen3.5 122B-A10B on Mac Studio M4 Ultra 256GB (256GB) — VRAM, KV Cache & Throughput

Qwen3.5 122B-A10B (122B total, 10B active, Hybrid Linear/Mamba + MoE) uses 69.19 GiB of 4-bit weights (Q4_K_M), requiring 70.03 GiB total at 8K context and 70.59 GiB at 32K on Mac Studio M4 Ultra 256GB (~64.9 tok/s decode).

Interactive Sizing Workbench

Full Compatibility Report →
256 GB · 819 GB/s
Fits with Safe Headroom
Config-derived formula HF config.json ↗
Memory Allocation Breakdown (GiB) 70.03 GiB (75.19 GB) / 256 GiB (27%)
Model Weights: 69.19 GiB KV Cache & State: 0.28 GiB Runtime & Activation Buffers: 0.56 GiB System RAM Offload: 0 GiB
Required GPU Memory
70.03 GiB
75.19 GB (decimal)
Usable Hardware Memory
199.7 GiB
256 GB (physical)
Safe Headroom
+129.67 GiB
27% utilization
Est. Decode Throughput
~64.9 tok/s
10B active params
Est. Prefill Throughput
~1596 tok/s
4K Prompt TTFT: ~2566 ms
Est. Cost / 1M Tokens
$1.89
Hybrid Linear/Mamba + MoE · Apache-2.0

Recommended Hardware for Current Configuration

Best Single GPU
RTX PRO 6000 Blackwell (96GB)
~$8,500
Best Value / Used GPU
2x A100 80GB SXM
160GB · ~$24,000
Apple Unified Memory
Ryzen AI Max+ 395 128GB
~$1,999
Cloud From
Vast.ai · $0.44 / hr
1x A100 80GB SXM

Production Launch Commands

llama.cpp (GGUF) 
llama-server -hf unsloth/Qwen3.5-122B-A10B-GGUF:Q4_K_M \
  --ctx-size 8192 -ngl 999 --flash-attn
vLLM 
vllm serve Qwen/Qwen3.5-122B-A10B \
  --max-model-len 8192 --gpu-memory-utilization 0.92
Live Token Speed Simulator (TTFT ~2566ms · ~64.9 tok/s)
[Qwen3.5 122B-A10B · 70.03 GiB] allocated on target GPU. Prefill completed in TTFT window; streaming autoregressive tokens at memory-bandwidth-bound rate...

Cloud GPU Comparison & Rent-vs-Buy TCO

Verified public rates across RunPod, Vast.ai, Vultr, and DigitalOcean.

Verified: 2026-11-04
Purchase break-even
~64.4 months
Est. local power: $12.77 / mo
Provider Recommended Instance Total VRAM Hourly Rate Monthly at Selected Hours Billing Mode Deploy
RunPod 1x A100 80GB SXM (Community Cloud · PCIe) 80 GB $1.19 / hr $144.70 / mo Per-second Deploy GPU ↗
Vast.ai 1x A100 80GB SXM (Verified Host · SXM4) 80 GB $0.44 / hr $53.75 / mo Spot / On-demand Deploy GPU ↗
Vultr ($300 Free Credit*) 1x A100 80GB SXM (Dedicated Instance · $300 Credit*) 80 GB $2.40 / hr $291.48 / mo Hourly Deploy GPU ↗
Vultr Cloud GPU 1x A100 80GB SXM (Dedicated Instance) 80 GB $2.40 / hr $291.48 / mo Hourly Deploy GPU ↗
DigitalOcean 1x Instinct MI300X 192GB (Dedicated Instance) 192 GB $2.59 / hr $314.94 / mo Per-second Deploy GPU ↗

Rates reflect public listings on the verification date and fluctuate with regional inventory.

Fine-Tuning VRAM Requirements (QLoRA / LoRA / Full) — Qwen3.5 122B-A10B

QLoRA (4b) Fits with Safe Headroom
71.63 GiB
Model Weights: 69.19 GiB · Gradients + Optimizer: 1.4 GiB · Act: 1.04 GiB
Recommended GPU: RTX PRO 6000 Blackwell (96GB)
LoRA (16b) Tight Fit (<8% Headroom)
230.83 GiB
Model Weights: 227.24 GiB · Gradients + Optimizer: 2.2 GiB · Act: 1.39 GiB
Recommended GPU: Instinct MI325X 256GB (256GB)
Full BF16 Exceeds Available GPU Memory
934.12 GiB
Model Weights: 227.24 GiB · Gradients + Optimizer: 704.44 GiB · Act: 2.44 GiB
Recommended GPU: 8x H200 141GB SXM (1128GB)

Engineering Sizing Notes & Optimization Playbook

Quick Hardware Verdict

Mac Studio M4 Ultra 256GB (256GB physical, 199.7 GiB usable) runs 4-bit (Q4_K_M) Qwen3.5 122B-A10B at 8K context using 70.03 GiB with 129.67 GiB of safe headroom (~64.9 tok/s single-stream decode).

KV Cache & Runtime Tuning

Qwen3.5 122B-A10B uses Hybrid Linear/Mamba + MoE (122B total, 10B active per token). Moving from 8K to 32K context expands FP16 KV cache from 0.28 GiB to 0.84 GiB; enabling 8-bit KV cache (llama.cpp -ctk q8_0 -ctv q8_0 or vLLM --kv-cache-dtype fp8) saves 0.38 GiB at 32K (reducing total VRAM to 70.21 GiB).

VRAM Scaling Across Context Windows (4K to 128K, 4-bit Weights)

Qwen3.5 122B-A10B VRAM scaling from 4K to 128K context (4K: 69.92 GiB, 8K: 70.03 GiB, 32K: 70.59 GiB, 128K: 72.84 GiB) 4K 69.92 GiB (KV 0.18 GiB) 8K 70.03 GiB (KV 0.28 GiB) 32K 70.59 GiB (KV 0.84 GiB) 128K 72.84 GiB (KV 3.09 GiB) Mac Studio M4 Ultra 256GB (256GB)
Model weights remain constant while attention KV cache grows with context length and batch size.

Quantization Precision & Memory Matrix — Qwen3.5 122B-A10B @ 8K

Quantization Format Weights (GiB) KV Cache Total VRAM (GiB) Decode Speed Compatibility (Mac Studio M4 Ultra 256GB) Recommended GPU Data Source
Native Weights (BF16 / FP8 / MXFP4 — Lossless) (BF16) 227.24 GiB 0.28 GiB 228.08 GiB (244.9 GB) ~22.1 tok/s Tight Fit (<8% Headroom) Instinct MI325X 256GB (256GB) Official checkpoint size
FP8 / GGUF Q8_0 (8-bit — Near-Lossless) (Q8_0) 120.72 GiB 0.28 GiB 121.56 GiB (130.52 GB) ~39.8 tok/s Fits with Safe Headroom Instinct MI300X 192GB (192GB) Config-derived formula
GGUF Q6_K (6-bit — High Precision) (Q6_K) 93.17 GiB 0.28 GiB 94.01 GiB (100.94 GB) ~50.1 tok/s Fits with Safe Headroom Instinct MI300X 192GB (192GB) Config-derived formula
GGUF Q5_K_M (5-bit — Balanced) (Q5_K_M) 80.97 GiB 0.28 GiB 81.81 GiB (87.84 GB) ~56.7 tok/s Fits with Safe Headroom RTX PRO 6000 Blackwell (96GB) Config-derived formula
Q4_K_M / UD-Q4_K_XL / AWQ (4-bit — Recommended) (Q4_K_M) 69.19 GiB 0.28 GiB 70.03 GiB (75.19 GB) ~64.9 tok/s Fits with Safe Headroom RTX PRO 6000 Blackwell (96GB) Config-derived formula
GGUF IQ4_XS (4.25-bit — Compact) (IQ4_XS) 60.58 GiB 0.28 GiB 61.42 GiB (65.95 GB) ~72.5 tok/s Fits with Safe Headroom RTX PRO 6000 Blackwell (96GB) Config-derived formula
GGUF Q3_K_M (3-bit — Low VRAM) (Q3_K_M) 55.73 GiB 0.28 GiB 56.57 GiB (60.74 GB) ~77.7 tok/s Fits with Safe Headroom RTX PRO 6000 Blackwell (96GB) Config-derived formula
GGUF UD-Q2_K_XL (2-bit — Extreme MoE) (UD-Q2_K_XL) 39.37 GiB 0.28 GiB 40.21 GiB (43.18 GB) ~102.1 tok/s Fits with Safe Headroom A40 48GB (48GB) Config-derived formula

Technical Specifications & FAQ

Does Qwen3.5 122B-A10B fit on Mac Studio M4 Ultra 256GB (256GB)?

In 4-bit (Q4_K_M) at 8,192 context, Qwen3.5 122B-A10B requires 70.03 GiB total GPU memory (69.19 GiB weights + 0.28 GiB KV cache + 0.56 GiB runtime buffers) against 199.7 GiB usable on Mac Studio M4 Ultra 256GB, delivering ~64.9 tok/s decode.

How much VRAM does Qwen3.5 122B-A10B need at 32K context?

At 32,768 context tokens, Qwen3.5 122B-A10B (Hybrid Linear/Mamba + MoE) allocates 0.84 GiB for KV cache and state, bringing total 4-bit VRAM to 70.59 GiB (75.8 GB decimal).

Evaluate This Model on Other GPUs

RTX 3060 12GB (12GB)RTX 4060 Ti 16GB (16GB)RTX 4070 Ti Super (16GB)RTX 4080 Super (16GB)RTX 5070 12GB (12GB)RTX 5070 Ti (16GB)RTX 5080 (16GB)RTX 3090 (24GB)RTX 4090 (24GB)RTX 4090 D (24GB)RTX 5090 (32GB)RTX 5090 D (32GB)RTX A6000 (48GB)RTX 6000 Ada (48GB)RTX PRO 6000 Blackwell (96GB)DGX Spark (GB10 128GB) (128GB)L4 24GB (24GB)A40 48GB (48GB)L20 48GB (48GB)L40S 48GB (48GB)A800 80GB SXM (80GB)A100 80GB SXM (80GB)H800 80GB SXM (80GB)H100 80GB SXM (80GB)H20 96GB HBM3 (96GB)H20 141GB HBM3e (141GB)H200 141GB SXM (141GB)B200 180GB (180GB)B300 288GB Blackwell Ultra (288GB)Ascend 910B 64GB (昇腾) (64GB)Ascend 910C 128GB (昇腾) (128GB)Atlas 300I Duo 96GB (310P) (96GB)MLU590 80GB (寒武纪思元) (80GB)DCU K100-AI 64GB (海光深算) (64GB)MetaX C500 64GB (沐曦曦云) (64GB)Kunlunxin P800 64GB (昆仑芯) (64GB)MTT S4000 48GB (摩尔线程) (48GB)Radeon RX 7900 XTX 24GB (24GB)Radeon AI PRO R9700 32GB (32GB)Ryzen AI Max+ 395 128GB (128GB)Instinct MI300X 192GB (192GB)Instinct MI325X 256GB (256GB)Instinct MI355X 288GB (288GB)Arc Pro B60 24GB (24GB)Gaudi 3 128GB HBM2e (128GB)Mac mini M4 Pro 64GB (64GB)MacBook Pro M4 Max 128GB (128GB)Mac Studio M3 Ultra 512GB (512GB)

Evaluate Other Models on This Hardware

Qwen3.8 Flash Next (125B-A6B)Qwen3.8 27BQwen3.8 2.4T-A95BQwen3.6 35B-A3BQwen3.5 397B-A17BQwen3.5 35B-A3BQwen3.5 27BQwen3.5 9BQwen3.5 4BQwen3 Coder 480B-A35BQwen3 Coder Next 80B-A3BQwen3 Coder 30B-A3BQwen3 30B-A3B Instruct 2507Qwen3 235B-A22B Instruct 2507Qwen3 32BQwen3 14BQwen3 8BQwen3 4B Instruct 2507Qwen3 1.7BQwen3 0.6BQwen2.5 72B InstructQwen2.5 Coder 32B InstructQwQ 32B ReasoningDeepSeek V4.1 Flash (552B-A16B)DeepSeek V4 FlashDeepSeek V3.2 (685B)DeepSeek R1-0528 671BDeepSeek V3-0324 671BDeepSeek R1 Distill Llama 70BDeepSeek R1 Distill Qwen 32BDeepSeek R1 Distill Qwen 14BDeepSeek R1 Distill Llama 8Bgpt-oss-20bgpt-oss-120bGemma 4 31B ITGemma 4 26B-A4B ITGemma 4 12B ITGemma 3 27B ITGemma 3 12B ITGemma 3 4B ITGLM-5.3 Flash (320B-A18B)GLM-5.3 (753B)GLM-5 (744B)GLM-4.5 (355B-A32B)GLM-4.5 Air (110B-A12B)GLM-4 32B-0414Kimi K3 (2.8T-A104B)Kimi K2 Instruct (1T-A32B)Kimi-Dev 72BMiniMax M3.1 Flash (428B-A23B)MiniMax M2.5 (229B-A10B)Step-3.7 Flash (198B-A11B)Step-3.5 Flash (197B-A11B)ERNIE 4.5 300B-A47B (文心)ERNIE 4.5 21B-A3B (文心)Hunyuan A13B Instruct (混元 80B)MiniCPM5 2B (面壁)Llama 4 Maverick 17B-128E (400B)Llama 4 Scout 17B-16E (109B)Llama 3.1 405B InstructLlama 3.3 70B InstructLlama 3.1 8B InstructLlama 3.2 3B InstructMistral Large 4 (1T-A52B)Mistral Large 2 123BMixtral 8x22B InstructMixtral 8x7B InstructMistral Small 3.2 24BDevstral Small 24BNemotron 3 Ultra 550B-A55BNemotron 3 Super 120B-A12BNemotron 3 Nano 30B-A3BCohere Command A 111BPhi-4 14BLiquid LFM3 1.7B