Pricing
The runtime is free. Models are one-time purchases - buy once, run forever on your hardware.
Free
$0
- Lean runtime (single binary, all platforms)
- lean-agent-35b (35B MoE, 3B active)
- lean-coder-80b (80B MoE, 3B active)
- All quantization variants
- OpenAI-compatible API server
Paid Models
One-time purchase
- lean-agent-122b (122B MoE, 10B active)
- lean-reason-397b (397B MoE, 17B active)
- lean-think-398b (398B afmoe, ~13B active)
- Full catalog bundle at a discount
- All quantization variants
Every model includes
- Model weights in
.lmpackformat with expert activation profiles - Multiple quantization variants (quality, balanced, compact)
- Download via
lean pull - Free updates to purchased models forever
- Commercial use - no restrictions on deployment
Hardware requirements
All models run via expert offloading. Download sizes shown at Q4_K_M quantization.
| Model | Q4 Size | Min VRAM | Min RAM | Price |
|---|---|---|---|---|
| lean-agent-35b | 21.4 GB | 12 GB | 16 GB | Free |
| lean-coder-80b | 48.7 GB | 12 GB | 32 GB | Free |
| lean-agent-122b | 75.0 GB | 24 GB | 32 GB | Coming soon |
| lean-reason-397b | 244.1 GB | 48 GB | 64 GB | Coming soon |
| lean-think-398b | 241.9 GB | 48 GB | 64 GB | Coming soon |