# Cactus

> Cactus is a cross-platform, open-source AI inference engine built in C/C++ to run language, speech, and vision models directly on smartphones, laptops, wearables, and other low-power edge hardware. It gives mobile developers Flutter, React Native, Kotlin, and Swift SDKs to deploy quantized models on-device - cutting latency to under 100ms, keeping data private, working offline, and avoiding cloud API bills - with automatic cloud fallback for heavier tasks. A YC Summer 2025 company, Cactus already powers production apps serving 500,000+ weekly inference tasks.

- **Founded:** 2025
- **Headquarters:** San Francisco, United States
- **Founders:** Roman Shemet (Co-Founder & CEO), Henry Ndubuaku (Co-Founder & CTO)
- **Team size:** ~8-27
- **Products:** Cactus Inference Engine, Cactus SDKs, Hybrid Cloud Fallback, Needle
- **Notable:** Accepted into Y Combinator Summer 2025., Powers production apps serving 500,000+ weekly inference tasks., Achieves 70+ tokens/sec on flagship phones (iPhone 17, Galaxy S25 Ultra) with NPU optimization.

## Products & services

- **Cactus Inference Engine** — A cross-platform, open-source AI runtime written in C/C++ that runs quantized language, speech, and vision models locally on phones, laptops, wearables, and edge devices. Supports GGUF models (Qwen, Gemma, Llama, DeepSeek, Phi, Mistral) from FP32 down to 2-bit, with zero-copy memory mapping for fast loading and low RAM.
- **Cactus SDKs** — Flutter, React Native, Kotlin/Java, and Swift/Objective-C bindings that let mobile developers embed on-device AI in an app with a few lines of code, including MCP tool-calls for device-native functions like reminders, gallery search, and messaging.
- **Hybrid Cloud Fallback** — Automatic routing that keeps simple or private tasks on-device and sends only complex tasks to the cloud, with an optional on-device-only mode so no data leaves the device.
- **Needle** — A 26M-parameter tool-calling model distilled from Gemini, small enough to run on constrained hardware such as wearables.

## Achievements

- Accepted into Y Combinator Summer 2025.
- Powers production apps serving 500,000+ weekly inference tasks.
- Achieves 70+ tokens/sec on flagship phones (iPhone 17, Galaxy S25 Ultra) with NPU optimization.
- 4,200+ GitHub stars on the open-source framework.
- Released Needle, a 26M-parameter tool-calling model distilled from Gemini.

## Latest updates

- **2025-08** — Joined Y Combinator's Summer 2025 batch.
- **2025-09** — Posted Launch HN: 'Cactus (YC S25) - AI inference on smartphones,' sharing benchmarks of 16-70 tokens/sec across phone tiers.
- **2025-09** — Reported 500,000+ weekly production inference tasks across deployments including AnythingLLM and KinAI.

## Links

- Website: https://cactuscompute.com
- LinkedIn: https://www.linkedin.com/company/cactus-compute
- Twitter/X: https://x.com/cactuscompute
- GitHub: https://github.com/cactus-compute/cactus

---

Profile page: https://yespress.io/cactus-yc-s25
Published by YesPress — https://yespress.io
Last updated: 2026-07-31