Skip to content
Run LLM Locally

Qwen3.6 35B-A3B on the RTX 3090

Qwen3.6 35B-A3B at Q4_K_M needs about 23 GB; the RTX 3090 has 24 GB: it fits with almost no margin at 8K context. At Q8_0 it does not fit in VRAM.

ShareFacebookXWhatsAppLinkedInEmail

Qwen3.6 35B-A3B

Q4_K_M · 8K context

Compatible with limits

The model needs about 23 GB (up to 23.9 GB) for 23.3 GB available: it fits, but with almost no margin. Use a shorter context or a smaller quantization if loading fails.

Your hardware
RTX 3090 24 GB · 64 GB RAM
Estimated memory
~23 GB estimate
VRAM (usable memory)
23.3 GB
GPU only
Yes
CPU/GPU offload
Not required
Generation speed
145–243 tok/s estimate
Largest context in fast memory
20K tokens
Confidence: low

Based on: memory bandwidth (936 GB/s) and model size

Memory breakdown

~23 GB (22.2 GB–23.9 GB) needed · estimate

Fits entirely in VRAM · 0.3 GB left

  • Model weights20.8 GB
  • KV cache0.2 GB
  • Runtime overhead1 GB
  • Safety margin1.1 GB
Assumptions: context 8K tokens · KV cache f16 · llama.cpp (CUDA) · flash attention on · weights = real file size

EstimateData updated 40 days agoSource: Run LLM Locally engine, Hugging Face config.json + model card

What makes this estimate less reliable

  • no benchmark for this hardware and model yet
  • mixture-of-experts: speed depends on expert routing

Other quantizations on the RTX 3090

Qwen3.6 35B-A3B on RTX 3090 + 64 GB RAM · 8K context
Qwen3.6 35B-A3B Q8_08K~38.2 GB · 14–27 tok/s estimateCPU/GPU offload
Qwen3.6 35B-A3B Q6_K8K~31.5 GB · 23–45 tok/s estimateCPU/GPU offload
Qwen3.6 35B-A3B Q5_K_M8K~26.6 GB · 44–87 tok/s estimateCPU/GPU offload
Qwen3.6 35B-A3B Q4_K_M8K~23 GB · 145–243 tok/s estimateCompatible with limits
Qwen3.6 35B-A3B IQ4_XS8K~20.5 GB · 153–260 tok/s estimateCompatible
Qwen3.6 35B-A3B Q3_K_M8K~18 GB · 162–280 tok/s estimateExcellent

Alternatives

GPUs that run Qwen3.6 35B-A3B Q4_K_M entirely in VRAM

Other models that fit the RTX 3090

Where to buy a RTX 3090

Prices and links follow the market you choose.

GeForce RTX 3090

24 GB GDDR6X · 936 GB/s · 350 W

Compatible with limitsRuns Qwen3.6 35B-A3B Q4_K_M entirely in memory

Price not available

Availability unknown

View on amazon.com

Affiliate link

No current offer imported for this market: the link opens a marketplace search.

Frequently asked questions

Can the RTX 3090 run Qwen3.6 35B-A3B?
The model needs about 23 GB (up to 23.9 GB) for 23.3 GB available: it fits, but with almost no margin. Use a shorter context or a smaller quantization if loading fails.
Which quantization should I use on the RTX 3090?
Q4_K_M is the highest-precision quantization that fits entirely in VRAM at 8K context. ~23 GB / 23.3 GB.
What is the largest context on the RTX 3090?
At Q4_K_M, about 20K tokens fit entirely in VRAM with an f16 KV cache. A q8_0 KV cache roughly doubles that.
How fast is Qwen3.6 35B-A3B (RTX 3090)?
We estimate 145–243 tok/s (estimate, confidence low). Based on: memory bandwidth (936 GB/s) and model size.