Features
Generates optimized CUDA/Triton kernels from any model
32 parallel Coder+Judge agents with MAP-Elites evolution
Pattern RAG with 1,711 CUTLASS and 113 Triton patterns
Up to 5× speedup over torch.compile(max-autotune) on Llama-3.1-8B
100% numerical correctness verification
Automatic Tensor Core optimization (WMMA, TMA) for Hopper
Three optimization modes: --turbo, default, --quality
Dual output: Triton Python kernels or native CUDA C++
Interactive CLI with wizard and benchmark browser
Session management for tracking past optimizations
Supports HuggingFace IDs, KernelBench tasks, or custom files
Credit system: 1 credit per kernel, 1-2 for HF models
Powered by fine-tuned Nemotron 3 Nano 30B at 250K tok/s
Drop-in replacement: same API, zero code changes
Laguna XS.2 model (33B params, 3B active, on-device)
Laguna M.1 model (225B params, 23B active, via API)
256K context length for multi-step reasoning
Multi-agent orchestration with planning and tool use
Sandboxed agent execution environments
Developer surfaces: agents, TUI, IDE extensions, binaries
Custom foundation models deployed on-prem or in VPC
Data connectors to repos, databases, data warehouses
Role-based access control for humans and agents
Executive-grade governance and auditability
On-prem, VPC, and workstation deployment (defense only)
Forward Deployed Research Engineers embed with teams
Real-time observability with end-to-end traces
Air-gapped network support
Open-weight model weights available
