Pricing
The runtime is free. Models are one-time purchases - buy once, run on your hardware.
Free
$0
- Lean runtime (single binary, Linux and Windows)
- lean-coder-welter (80B MoE, 3B active)
- lean-agent-light (35B MoE, 3B active)
- One
.lmpackper model - OpenAI-compatible API server
Paid Models
Launch pricingOne-time purchase
- lean-agent-middle (122B MoE, 10B active) - checkout opening soon$19.99
- lean-agent-cruiser (284B MoE, 13B active) - coming soon
- lean-think-heavy (398B afmoe, ~13B active) - coming later
- lean-reason-heavy (397B MoE, 17B active) - coming later
- One
.lmpackper model - Free updates to purchased models
Checkout is opening soon.
Launch pricing - prices may change after the launch window. Purchases are one-time; updates to purchased models stay free.
Every model includes
- Model weights in
.lmpackformat with expert activation profiles - A single
.lmpackper model - the quantization it is built from varies by model and is listed below - Download via
lean pull - Output verified against llama.cpp on identical weights
- Commercial use - no restrictions on deployment
Hardware requirements
All models run via expert offloading. Download sizes vary by quantization level.
lean-agent-middle
VRAM: 24 GBRAM: 32 GBPaid
Download 77.67 GB - single .lmpack, from Q4_K_M
lean-coder-welter
VRAM: 12 GBRAM: 32 GBFree
Download 49.72 GB - single .lmpack, from UD-Q4_K_XL
lean-agent-light
VRAM: 12 GBRAM: 16 GBFree
Download 22.06 GB - single .lmpack, from Q4_K_M
Not yet released
Target specs for the unreleased models, pending validation on our hardware.
lean-agent-cruiser
Target VRAM: 24 GBRAM: 64 GBComing soon
Download 155.14 GB - single .lmpack, from UD-Q4_K_XL
lean-think-heavy
Target VRAM: 48 GBRAM: 64 GBComing later
Download ~241.85 GB - single .lmpack, from Q4_K_M (target)
lean-reason-heavy
Target VRAM: 48 GBRAM: 64 GBComing later
Download 226.17 GB - single .lmpack, from UD-Q4_K_XL