LOCAL_AI_STACK
v2.2 · 2026-06-12
← back to models
VISION / MULTIMODAL MODEL — RUNTIME REQUIREMENTS
·Beginner summary from supplied research: A high-end MoE Qwen3.6 option for users who want the newer Qwen multimodal agent stack but prefer an active-parameter MoE profile over a dense 27B model.

Qwen3.6 35B-A3B

qwen3-6-35b-a3b
MULTIMODAL

Qwen3.6 35B-A3B is a 35B multimodal model profile for local AI planning. This page records provider, license, context, quantization, hardware-fit estimates, setup hints, source links, and caveats so readers do not choose a model by name alone.

PROVIDER
Alibaba Qwen Team
FAMILY
Qwen3.6
MODEL TYPE
multimodal
PARAMETERS
35B (3B active)
MODALITIES
text, image
ARCHITECTURE
Sparse MoE hybrid Gated DeltaNet / FFN / Gated Attention model with 256 experts, 8 routed experts, 1 shared expert, and vision encoder
CONTEXT WINDOW
262,144 tokens
TRAINING TOKENS
—
RELEASE DATE
2026-04-16
LICENSE & LINKS
SETUP HINTS
OLLAMA
ollama run qwen3.6:35b-a3b
LM STUDIO
Available according to the supplied June 2026 research; select a cited GGUF/MLX build and verify current app metadata before relying on hands-on setup guidance.
LLAMA.CPP
Use a cited GGUF build with current llama.cpp-compatible tooling.
QUANTIZATION FILE SIZES (GGUF)
Q4_K_M
24 GB
Q8_0
39 GB
Quantization note: Q4_K_M for first local attempts unless a cited runtime recommends a different default.
RAM / VRAM ESTIMATES
MIN RAM
32 GB
COMFORTABLE RAM
64 GB
MIN VRAM
32 GB
COMFORTABLE VRAM
64 GB
These are conservative local-inference estimates, not official hardware requirements.
HARDWARE FIT
CPU only (no GPU)
no
8 GB RAM
no
16 GB RAM
no
32 GB RAM
limited
8 GB VRAM
no
12 GB VRAM
no
24 GB VRAM
no
Apple Silicon (unified memory)
limited
Hardware fit values are conservative local-inference estimates based on GGUF size plus runtime and KV-cache overhead. Actual requirements depend on context length, quantization, runtime, GPU offload, and other running apps.
BEST FOR
·High-memory MoE testing, local agents, coding, and multimodal workflows.
AVOID IF
·You expect A3B active parameters to mean tiny memory use.
CAVEATS
·A3B means about 3B active parameters per token, not a 3B download or 3GB memory footprint.
·The Q4 model is larger than Qwen3.6 27B Q4.
·For most beginners, Qwen3.5 9B is the better first recommendation.
·Commercial-use status from the supplied research: permitted
·Beginner summary from supplied research: A high-end MoE Qwen3.6 option for users who want the newer Qwen multimodal agent stack but prefer an active-parameter MoE profile over a dense 27B model.
SOURCE URLS
FIELD EVIDENCE
hfGgufRepos—
q5FileSizeGb—
trainingTokens—
LANGUAGES
Unknown
CAPABILITIES
tool callingreasoning tunedvision / image input
← Qwen3.6 27BGPT-OSS 20B →