Maker / Hobbyist
For learning, prototypes and side projects.
$0/mo
Free forever
- 50 model compilations / month
- Basic INT8 quantization
- Targets: ESP32 & Cortex-M4
- Community support
- Standard C++ headers
Pricing
Pay for what you compile
Start free. Scale to production. Go custom for automotive-grade deployments.
Monthly
Pay AnnuallySave 17%
Pro Engineer
For engineers shipping edge-AI to production hardware.
$29/mo
Billed monthly
- 2,000 model compilations / month
- Advanced quantization & magnitude pruning
- All targets: STM32, ESP32, RISC-V, ARM Cortex-M/R, NXP
- Bare-metal zero-dependency C++ export
- Priority email & Discord support
Automotive / Enterprise
For OEMs and Tier-1 suppliers with safety-critical needs.
Custom
Talk to engineering
- Unlimited compilations & on-premises CLI
- CAN-Bus / AUTOSAR integration & MISRA-C compliance
- Custom MCU architecture & NPU acceleration targets
- Dedicated Edge-AI solutions engineer
- SLA guarantees
How many compilations do you need?
200 compilations / month
Recommended: Pro Engineer
101001,00010,000+
Compare plans
| Feature | Maker | Pro Engineer | Enterprise |
|---|---|---|---|
| Compilation limits & rates | |||
| Compilations / month | 50 | 2,000 | Unlimited |
| Processing queue | Standard | Priority | Dedicated |
| On-premises CLI | — | — | ✓ |
| Optimization features | |||
| FP32 → INT8 quantization | Basic | Advanced | Advanced + custom |
| Magnitude pruning | — | ✓ | ✓ |
| Memory footprint profiler | — | ✓ | ✓ |
| Hardware & toolchain support | |||
| Targets | ESP32, Cortex-M4 | STM32, ESP32, RISC-V, Cortex-M/R, NXP | All + custom MCU / NPU |
| GCC / Clang | ✓ | ✓ | ✓ |
| Keil / STM32Cube | — | ✓ | ✓ |
| CAN-Bus / AUTOSAR · MISRA-C | — | — | ✓ |
| Security & privacy | |||
| Model encryption | — | ✓ | ✓ |
| On-premises execution | — | — | ✓ |
| Support | Community | Email & Discord | Dedicated engineer + SLA |
Frequently asked questions
Do my trained model weights leave my local machine?
On the hosted compiler, your model file is uploaded to be compiled and is not shared. If weights must never leave your network, the Enterprise on-premises CLI runs the whole pipeline locally.
Are the generated C++ headers free of runtime dependencies?
Yes. Output is plain C++ with a static memory arena: no heap allocation, no external inference runtime, and no third-party libraries to link.
How do compilation credits/tokens work?
Each compilation uses tokens from your plan's monthly allowance. Your remaining balance is always shown in the navbar once you're signed in, and it resets each billing cycle.
Can I request custom hardware target support?
Yes. Custom MCU architectures and NPU targets are part of the Enterprise plan. Contact engineering with your chip and toolchain and we'll scope it with you.