ai.canshu.com
中文
VRAM Calculator Advisor & TCO API Pricing Compare Models Leaderboard GPUs

Search models, GPUs, and tools

↑↓ Navigate · Enter Open · ★ Watchlist prioritized ai.canshu.com
Q4_K_M: Smooth Fit 23.96 GB / 46.1 GB Usable

Qwen3.6-35B-A3B (MoE) on 摩尔线程 MTT S4000 48GB (GDDR6): VRAM Requirement & Speed Estimate

Yes. At Q4_K_M 4-bit quantization with a 16K context window, Qwen3.6-35B-A3B (MoE) requires 23.96 GB of VRAM (21.83 GB weights + 0.4 GB KV cache + 1.74 GB overhead), fitting within 摩尔线程 S4000 48G's 46.1 GB usable memory at ~273.8 tokens/s.

Concurrency (Batch):
MoE 36B / A3B·40L (d=2048)·16Q/2KV·Ctx 256K·Text+Image+Video
Runnable Models →
GPUs
48 GB VRAM·768 GB/s·Safe Usable 46.1 GB
Smooth Fit23.96 / 46.1 GB~273.8 tok/s
3. Weight Quantization
4. Per-Request Context16K (16,384 tok)
Verified: 2026-11-20·Roofline bandwidth/compute estimate (Batch=1)
Smooth Fit
23.96/ 46.1 GB usable

Physical 48 GB · Load 52%

Theoretical Decode Speed
~273.8tokens/s

MoE: reads only 1.82 GB active weights/step

Max Context at Batch=1
256K/ 256K

Enable Q8/Q4 KV quantization to scale batch concurrency

Weights: 21.83 GBKV Cache (1×16K): 0.4 GBOverhead: 1.74 GB
🚀 Blazing Fast (Instant CoT)~1.8s for a 500-token response
Reference Launch CLI:
OLLAMA_CONTEXT_LENGTH=16384 ollama run qwen3-6-35b-a3b

摩尔线程 S4000 48G Fits Qwen3.6-35B-A3B (MoE) · Cost & TCO

Approx. $5,000 (~$104/GB VRAM) · 450W TDP

📊 All Quantization Levels for Qwen3.6-35B-A3B (MoE) on 摩尔线程 S4000 48G

Quantization Total VRAM (16K) Verdict Est. Speed Max Context
FP16 / BF16 75.14 GB Out of VRAM (OOM) — —
FP8 / Q8_0 40.72 GB Tight Fit (Watch Context) ~150.1 tok/s 256K
GGUF Q6_K 31.81 GB Smooth Fit ~207.8 tok/s 256K
GGUF Q5_K_M 27.82 GB Smooth Fit ~236.9 tok/s 256K
Q4_K_M / AWQ 23.96 GB Smooth Fit ~273.8 tok/s 256K
GGUF IQ4_XS 21.21 GB Smooth Fit ~308.1 tok/s 256K
GGUF Q3_K_M 19.65 GB Smooth Fit ~331.7 tok/s 256K
GGUF Q2_K 16.16 GB Smooth Fit ~400.1 tok/s 256K

💡 Recommended Alternative GPUs & Macs for Qwen3.6-35B-A3B (MoE)

Frequently Asked Questions

Can 摩尔线程 MTT S4000 48GB (GDDR6) run Qwen3.6-35B-A3B (MoE) locally?

Yes. At Q4_K_M 4-bit quantization with a 16K context window, Qwen3.6-35B-A3B (MoE) requires 23.96 GB of VRAM (21.83 GB weights + 0.4 GB KV cache + 1.74 GB overhead), fitting within 摩尔线程 S4000 48G's 46.1 GB usable memory at ~273.8 tokens/s.

What is the best quantization for Qwen3.6-35B-A3B (MoE) on 摩尔线程 S4000 48G?

The highest-quality quantization that fits cleanly on 摩尔线程 S4000 48G is FP8 / Q8_0 (99.9% Near-Lossless), using 40.72 GB VRAM at ~150.1 tokens/s with up to 256K context.

How much VRAM does Qwen3.6-35B-A3B (MoE) need in full FP16 precision?

In unquantized FP16/BF16 precision (16K context), Qwen3.6-35B-A3B (MoE) requires 75.14 GB VRAM (72 GB for weights alone). Because it uses a Mixture-of-Experts (MoE) architecture, all 36B parameters must sit in memory, while only 3B active parameters are read per token.