Qwen3.5 122B A10B DFlash — Hardware Requirements & GPU Compatibility
Chat
Qwen3.5 122B A10B DFlash is a 122B-parameter open language model from z-lab in the Qwen 3.5 family. It supports a context window of up to 262,144 tokens. At Q4_K_M it needs about 73.54 GB of VRAM — see which GPUs and Macs can run it below.
Qwen3.5 122B A10B DFlash runs on a 96 GB+ Mac or a workstation GPU
Best local pick → Mac Studio (M3 Ultra, 96GB) · Q4_K_M · 73.8 GB · ~85 tok/s
Macs from 64 GB (Q3_K_M · ~101 tok/s)
Apache 2.0
Not set up for it?
Reference GPU · RTX 4090 24 GB
Estimates, ±20% · how we estimate
Run Qwen3.5 122B A10B DFlash
One recommended way. Every other method, OS and quant is in the step-by-step guide.
Recommended
vllm serve z-lab/Qwen3.5-122B-A10B-DFlash --max-model-len 16384vLLM repo with its own quantisation
Download: Weights on Hugging Face
Other ways
each opens step-by-stepRun in cloud
Fits on RTX PRO 6000 (96 GB) (22 GB headroom) · Q4_K_M
- Generation speed
- ~182 tok/s
- generation speed
- Cost per 1M output tokens
- $2.04
- per 1M output tokens
How Much VRAM Does Qwen3.5 122B A10B DFlash Need?
Select a quantization level; the listed hardware adheres to it.
| Quantization | VRAM | +32K | On RTX 4090 (reference) | |
|---|---|---|---|---|
| 52.2 GB | 52.8 GB | 2-bit · K-quant — 2-bit quantization with K-quant improvements | Does not fit: needs 52.5 GB | |
| 59.8 GB | 60.4 GB | 3-bit medium — 3-bit medium quantization | Does not fit: needs 60.1 GB | |
| 73.5 GB | 74.1 GB | 4-bit medium — 4-bit medium quantization — most popular sweet spot | Does not fit: needs 73.8 GB | |
| 87.3 GB | 87.8 GB | 5-bit medium — 5-bit medium quantization — good quality/size tradeoff | Does not fit: needs 87.5 GB | |
| 101.0 GB | 101.6 GB | 6-bit, very good — 6-bit quantization, very good quality | Does not fit: needs 101.3 GB | |
| 122.3 GB | 122.9 GB | 8-bit, near-lossless — 8-bit quantization, near-lossless | Does not fit: needs 122.6 GB | |
| 244.3 GB | 244.9 GB | 16-bit · bfloat — Brain floating point 16 — preferred for training | Does not fit: needs 244.6 GB |
Selected Q4_K_M
4-bit medium quantization73.8GBat 16K context
- Weights
- 73.2 GB
- KV cache
- 0.3 GB
- Overhead
- 0.3 GB
Doesn't fit RTX 4090 (reference) · needs 49.8 GB more
Which Hardware Can Run Qwen3.5 122B A10B DFlash?
Q4_K_M · 73.5 GBQwen3.5 122B A10B DFlash (Q4_K_M) requires 73.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 96+ GB is recommended. Using the full 262K context window can add up to 4.8 GB, bringing total usage to 78.3 GB. No consumer GPU has enough memory. 30 devices with unified memory can run Qwen3.5 122B A10B DFlash, including Mac Studio (M3 Ultra, 256GB), Mac Studio (M3 Ultra, 512GB), Mac Studio (M3 Ultra, 96GB).
64–80 GB
Q4_K_M · best fit Q3_K_M (60.1 GB)
3 Mac classes · 1 device family
M2 Ultra*64 GB~101Simulate output on Mac Pro M2 Ultra (64 GB)M5 Max64 GB~79Simulate output on Mac Studio M5 Max (64 GB)Orin~27Simulate output on NVIDIA Jetson AGX Orin 64GB96 GB
fits · 22.2 GB free · up to 256K ctx
2 Mac classes
M3 Ultra*96 GB~85Simulate output on Mac Studio (M3 Ultra, 96GB)M3 Max*96 GB~33Simulate output on MacBook Pro 14" M3 Max (96 GB)
By VRAM tier · typical GPUs and Macs at Q4_K_M
Alternatives to Qwen3.5 122B A10B DFlash
Qwen 3.5 family · 34 models →Similar models
6| Model | Score | VRAM | tok/s | Why it is similar |
|---|---|---|---|---|
| GPT OSS 120BOpenAI | 97.5 | 70.8 GB | —needs > 24 GB | same size |
| Ornith 1.5 122Bornith-ai | — | 75.5 GB | —needs > 24 GB | same size |
| NVIDIA Nemotron 3.5 Super VL 120B A12B BF16NVIDIA | — | 87.7 GB | —needs > 24 GB | same size |
| Ling 3.0 Flash VLInclusion AI | — | 76 GB | —needs > 24 GB | same size |
| Leanstral 1.5 119B A6BMistral AI | — | 84 GB | —needs > 24 GB | same size |
| Laguna S 2.1poolside | — | 72.5 GB | —needs > 24 GB | same size |
Specifications
- Publisher
- z-lab
- Family
- Qwen 3.5
- Parameters
- 122B
- Active parameters
- 10B per token
- Architecture
- DFlashDraftModel
- Context Length
- 262,144 tokens
- Vocabulary Size
- 248,320
- Release Date
- 2026-04-21
- License
- Apache 2.0
Frequently Asked Questions
How much VRAM does Qwen3.5 122B A10B DFlash need?
Qwen3.5 122B A10B DFlash requires 73.5 GB of VRAM at Q4_K_M, or 244.3 GB at BF16. Full 262K context adds up to 4.8 GB (78.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 122B × 4.8 bits ÷ 8 = 73.2 GB
KV Cache + Overhead ≈ 0.3 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 ≈ 5.1 GB (at full 262K context)
VRAM usage by quantization
Can NVIDIA GeForce RTX 5090 run Qwen3.5 122B A10B DFlash?
No — Qwen3.5 122B A10B DFlash requires at least 52.2 GB at Q2_K, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.
What's the best quantization for Qwen3.5 122B A10B DFlash?
For Qwen3.5 122B A10B DFlash, Q4_K_M (73.5 GB) offers the best balance of quality and VRAM usage. Q5_K_M (87.3 GB) provides better quality if you have the VRAM. The smallest option is Q2_K at 52.2 GB.
VRAM requirement by quantization
★ Recommended — best balance of quality and VRAM usage.
Can I run Qwen3.5 122B A10B DFlash on a Mac?
Yes — Mac Pro M2 Ultra (64 GB) and 41 other Macs can run Qwen3.5 122B A10B DFlash, the smallest at Q3_K_M. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 73.5 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 96GB).
Can I run Qwen3.5 122B A10B DFlash locally?
Yes — Qwen3.5 122B A10B DFlash can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.
How fast is Qwen3.5 122B A10B DFlash?
At Q4_K_M, Qwen3.5 122B A10B DFlash can reach ~407 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 B200 → 6.0 GB active ÷ (8000 × 0.65) = 1.159 ms, plus 6 layers × 0.055 ms = 0.330 ms, so 1000 ÷ 1.489 ms = ~672 tok/s
Estimated speed at Q4_K_M (73.5 GB)
Real-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
What's the download size of Qwen3.5 122B A10B DFlash?
At Q4_K_M, the download is about 73.20 GB. The full-precision BF16 version is 244.00 GB. The smallest option (Q2_K) is 51.85 GB.
Which GPUs can run Qwen3.5 122B A10B DFlash?
No single consumer GPU has enough VRAM to run Qwen3.5 122B A10B DFlash at Q4_K_M (73.5 GB). Multi-GPU or professional hardware is required.
Which devices can run Qwen3.5 122B A10B DFlash?
30 devices with unified memory can run Qwen3.5 122B A10B DFlash at Q4_K_M (73.5 GB), including ASUS Ascent GX10, Beelink GTR9 Pro (Ryzen AI Max+ 395, 128 GB), Framework Desktop (Ryzen AI Max+ 395, 128 GB), GMKtec EVO-X2 (Ryzen AI Max+ 395, 128 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.