Back to registry
CUDA16GB VRAM

arnav080/qwen3.6-35b-moe-rtx5080-128k-LeftCDev

Fully GPU-accelerated Qwen 3.6 35B MoE recipe optimized for 16GB VRAM GPUs, achieving 150 tok/s and 128k context.

Configuration Specifications

Base Modelhuggingface:Qwen/Qwen3.6-35B-A3B-Instruct
Enginellama.cpp
QuantizationQ3_K_S
Platformcuda
VRAM Required16GB minimum

Telemetry Benchmarks

No runs recorded yet

Be the first to benchmark this recipe! Run the CLI command from your terminal:

bloc run arnav080/qwen3.6-35b-moe-rtx5080-128k-LeftCDev
qwen3.6-35b-moe-rtx5080-128k-LeftCDev.yaml
Loading workspace editor...