tinyanalysis

Leaderboard / Alibaba (Qwen)

Qwen3.6 35B-A3B

Alibaba (Qwen) · Qwen3.6 · released 2026-04-16 · data checked 2026-11-01

Qwen3.6 35B-A3B is a 35B-parameter Mixture-of-Experts model (3B active per token) from Alibaba (Qwen) (China), released 2026-04-16 under the Apache-2.0 license. It scores 72.2 on the Small LM Index, #6 of 34 medium models. It needs about 22.6 GB of memory at 4-bit with an 8K-token context and costs $0.031 per million tokens through the cheapest API provider.

Small LM Index
72.2
#6 of 34 scored medium models
Likely 70.9–73.6 · from 6 of 6 core benchmarks
Parameters
35B
3B active · MoE
Price per 1M tokens
$0.031
$0.010 in / $0.092 out · blended 3:1
Context window
262K
tokens

Benchmarks

Each bar shows this model's score. The tick marks the best medium model on that benchmark. Sources: 8 reported by Alibaba (Qwen). How we label sources.

MMLU-Pro Lab-reported Knowledge85.2
Class best: Qwen3.6 27B 86.2
GPQA Diamond Lab-reported Knowledge86.0
Class best: Nanbeige4.3 33B 89.5
Humanity's Last Exam Lab-reported Reasoning21.4
Class best: Qwen3.8 27B 30.8
AIME Lab-reported Math · AIME 202692.7
Class best: Ternary Bonsai 2 27B 95.8
LiveCodeBench Lab-reported Coding · LiveCodeBench v680.4
Class best: Qwen3.8 27B 90.3
SWE-bench Verified Lab-reported Coding73.4
Class best: Qwen3.6 27B 77.2
Terminal-Bench Lab-reported Agentic · Terminal-Bench 2.051.5
Class best: Nanbeige4.3 33B 80.9
MMMU-Pro Lab-reported Vision75.3
Class best: Gemma 4 31B 76.9

Specifications

Lab
Alibaba (Qwen) (CN · China)
Architecture
Mixture-of-Experts
Total parameters
35B
Active parameters
3B
Size class
Medium
Context window
262K tokens
Input modalities
text, image, video
Reasoning
Hybrid (thinking on/off)
License
Apache-2.0
Openness
Open weights
Memory to run
~22.6 GB at 4-bit, 8K context (more sizes)
Hugging Face
4.5M downloads (30 days) · 3.1K likes
API providers
10 on OpenRouter
API price
$0.010 in / $0.092 out per 1M (OpenRouter, 2026-11-03)
Weights on Hugging Face ↗Try on OpenRouter ↗

Compare with

Ranked in

Sources

Something wrong here? Report it.

Run it locally

Uses ggml-org/Qwen3.6-35B-A3B-GGUF (Q4_K_M). Swap the tag for another size: BF16, Q4_0, Q4_K_M, Q8_0.

Ollama
ollama run hf.co/ggml-org/Qwen3.6-35B-A3B-GGUF:Q4_K_M
llama.cpp
llama-server -hf ggml-org/Qwen3.6-35B-A3B-GGUF:Q4_K_M
Then open http://localhost:8080 for a chat window.
MLX (Mac)
pip install -U mlx-lm
mlx_lm.chat --model mlx-community/Qwen3.6-35B-A3B-4bit
Text only; for images use mlx-vlm.
vLLM (GPU server)
vllm serve Qwen/Qwen3.6-35B-A3B
Full-precision weights, for a server GPU.

Open in LM Studio

Memory needed: weights, context cache and runtime buffers.

Quantization4K context32K context256K context
16-bit F1671 GB72 GB77 GB
8-bit Q8_038 GB39 GB44 GB
4-bit Q4_K_M23 GB23 GB28 GB

Check it against your GPU, Mac or phone →