Alibaba
Qwen 3.6 35B A3B
最先端
5.7Mダウンロード2.0KいいねApr 2026公開日262K トークンコンテキストApache 2.0ライセンス98 卓越品質
Qwen 3.6 35B A3B (35B parameters) requires approximately 28.5 GB of VRAM with Q4_K_M quantization. As a Mixture of Experts model with 3B active parameters, it uses less memory than its total parameter count suggests. For the best balance of quality and speed, we recommend hardware with at least 33 GB of VRAM.
はじめに
— コピー&ペーストでローカル実行Copy-paste commands to run Qwen 3.6 35B A3B on your machine.
Run
docker run --rm -it ghcr.io/ggerganov/llama.cpp:full \
--hf-repo "Qwen/Qwen3.6-35B-A3B" \
--hf-file "Qwen3.6-35B-A3B-Q4_K_M.gguf" \
-c 4096 -ngl 99Quick specs
Parameters35B (3B active)
Architecturemoe (MoE)
Context262K tokens
Modalitytext+vision
Min RAM13.7 GB
Rec. RAM21.3 GB (Q4_K_M)
LicenseApache 2.0
FamilyQwen
✓ Vision✓ Code✓ Chat✓ Reasoning
About this model
- •35B total params with only 3B active per token
- •262K native context with preserve-thinking support
- •Multimodal open-weights model tuned for coding and agent workflows
関連モデル
おすすめ
最適なハードウェア
Qwen 3.6 35B A3Bのおすすめ
このモデルを実行
量子化オプション
量子化レベル別VRAM推定値
No hardware detected — fit column shows raw VRAM estimates
| Quant | Bits | VRAM | Quality | Fit |
|---|---|---|---|---|
Q2_K | 2 | 13.7 GB | Low | — |
Q3_K_S | 3 | 17.2 GB | Low | — |
NVFP4 | 4 | 19.6 GB | Medium | — |
Q4_K_M | 4 | 21.3 GB | Medium | — |
Q5_K_M | 5 | 25.2 GB | High | — |
Q6_K | 6 | 28.7 GB | High | — |
Q8_0 | 8 | 37.5 GB | Very High | — |
F16 | 16 | 71.8 GB | Maximum | — |
Quality benchmarks
Qwen 3.6 35B A3B benchmark scores
Coding
SWE-bench Verified73.4%
HumanEval+—
Aider Polyglot—
LiveCodeBench80.4%
Reasoning
MMLU-Pro85.2%
GPQA Diamond86.0%
MATH-500—
ARC Challenge—
Source: official · 2026-04-15
ハードウェア互換性
全ハードウェアの適合度推定
Computing compatibility...
メモリ内訳
Reference: RTX 2060 6GB
Weights21.3 GB
KV Cache4.1 GB
Runtime2.4 GB
Headroom0.6 GB
よくある質問
FAQ — Qwen 3.6 35B A3B
関連項目