AI Train2026
Runs transformer forward+backward training directly on Apple's Neural Engine by reverse-engineering private ANE APIs; includes per-layer ANE kernels, INT8 optimizations and benchmarks. Proof-of-concept only—relies on undocumented APIs, has low utilization and CPU fallbacks.
AI Audio2026
Runs local speech-to-text inference for a wide range of ASR model families using GGUF models on the ggml runtime. Supports Metal, Vulkan, and CUDA GPU backends plus a tinyBLAS-accelerated CPU path, prebuilt GGUFs on Hugging Face, and a quantization tool.
AI Deploy2026
Runs local LLMs on Apple Silicon using native MTP speculative decoding to accelerate token generation while preserving the model's output distribution. Leverages the model's own MTP heads with batched verification and exact rejection sampling; ships with a Mac app, CLI, local OpenAI/Anthropic-compatible server, auto-tune, and Forge for building/verifying MTP adapters.
AI Infra2026
Native local inference engine for DeepSeek V4 Flash (also supports GLM 5.2 and PRO on high‑memory machines). Focused features include model-specific loading, SSD expert streaming, asymmetric routed-expert 2-bit quant support, multi-GPU/tensor/pipeline parallelism, and an OpenAI-compatible server plus a native coding agent.
AI Agent2026
Build native desktop apps authored with declarative .native markup and TypeScript (or Zig) compiled to native code, with no browser or JS runtime in the binary. Ships a component catalog, deterministic rendering, hot reload, and an embedded automation server for AI agent workflows.
AI Infra2026
Streams routed experts from disk and treats storage, RAM and VRAM as a single inference hierarchy so MoE models (hundreds of billions to trillions of parameters) can run on consumer or heterogeneous hardware; single-file C engines with zero runtime deps and optional Vulkan/CUDA/Metal backends, speed governed by placement and disk I/O.
AI Model2026
Runs a full 27B-class language model using end-to-end binary (1.125-bit) weights, cutting FP16 size to ~3.9 GB. Key features: 262k-token context, custom 1-bit kernels for Apple MLX and CUDA, and an optional DSpark drafter for faster decoding. Best when memory footprint matters; trades some FP16 accuracy for on-device feasibility.
AI Model2026
Runs a full 27B-class Qwen3.6-derived LLM in a ~7.2 GB ternary/2‑bit format for on-device or single‑GPU text generation, retaining ~95% of FP16 performance and supporting a 262K‑token context. Designed for laptop/GPU deployment; exceeds typical phone memory limits.
AI Model2026
Provides a quantized MLX conversion of Qwen3.8-27B for Apple Silicon (2/4/6/8-bit) with the model's refusal-direction ablated, preserving multimodal vision+text capability; intended for red‑teaming, interpretability and safety research, not unmoderated production use.
AI Model2026
Compresses a 27B-class multimodal model into end-to-end ternary weights to run 27B reasoning on-device: 5.9–8.6 GB deployed footprint, 262K-token context, ~98.2% of FP16 benchmark performance; ships MLX and GGUF packs and runs on Apple MLX and CUDA.
AI Model2026
A 27B-class language model packaged in GGUF with end-to-end ternary weights for on-device or single-GPU llama.cpp inference; reduces FP16 footprint to ~5.9–7.2 GB while retaining ~98% of baseline performance and supporting up to 262K tokens.
AI Deploy2026
Evaluates multi-field JSON schemas in parallel to extract boolean or categorical field values from text, producing guaranteed-valid JSON and per-field calibrated confidences. Uses KV-cache broadcasting, sub-vocabulary logit slicing and token-tree disambiguation to cut latency (5.6x–7.0x on M4 Max) versus autoregressive decoding; requires Apple Silicon and MLX.