lean-agent-light
Lightweight class · v1.0 · Qwen 3.5 base
General-purpose agent - tool calling, structured output, multi-step reasoning
Verified vs llama.cppFree
The entry point. A 21 GB model (Q4_K_M) that runs on 12 GB VRAM - expert offloading handles the rest. Qwen3.5 GDN hybrid architecture surpasses last-gen models many times its size.
Specifications
Total params
35B
Active per token
3B
Base model
Qwen3.5-35B-A3B
Architecture
GDN hybrid MoE
Experts
256 experts, 8 active
Min VRAM / RAM
12 GB / 16 GB
Measured performance
Real numbers from our reference rig, not projections.
| Hardware | Prefill | Decode |
|---|---|---|
| 1x RTX 3090 (24 GB) | 14-31 tok/s | 19-31 tok/s |
| 2x RTX 3090 | 10-23 tok/s | 15-27 tok/s |
Download
Download 21.4 GB - single .lmpack, Q4_K_M
$ lean pull lean-agent-lightQuick start (coming soon)
$ curl -sSf https://leanmodels.ai/install.sh | sh
$ lean pull lean-agent-light
$ lean run lean-agent-lightSingle binary, 15 MB. No Python, no Docker, no cloud dependency.
The installer is not live yet — binaries ship with our first release.