| Platform | Apple Silicon macOS 14 or later | macOS, Windows, and Linux | macOS, Windows, and Linux | Apple Silicon macOS, macOS 15.0+ | Apple Silicon macOS | Apple Silicon macOS |
|---|
| Core runtime | Swift in-process inference through Apple MLX; the default path requires no Python runtime | Ollama service with platform model runners; official MLX support on Apple Silicon, with structured output in the MLX runner as of v0.33.1 | Multiple runtimes; the official MLX engine is Python-based and bundles its own Python | Python/FastAPI inference core with native SwiftUI shell | Pure Swift over MLX | Native Swift MLX server; no Python core documented |
|---|
| Model workflow | Supported MLX language, vision, embedding, LoRA, and checkpoint-governed native model workflows | Ollama model library and model import workflow | Managed discovery and download workflows across supported formats | Multi-model MLX serving | Language, vision, embedding, and audio model workflows | MLX language models with very large MoE focus |
|---|
| Interfaces | SwiftUI app, macmlx CLI, compatible HTTP APIs with structured output, and integrated tool-routing surfaces | App, CLI, native Ollama API, and selected OpenAI-compatible routes | LM Studio app, llmster headless daemon, lms CLI, SDKs, and local APIs | Menu-bar app, server tools, OpenAI and Anthropic compatibility | Menu-bar app, CLI, model management, and OpenAI-compatible APIs | OpenAI-compatible server and SwiftBuddy GUI |
|---|
| Factual focus / audience | Swift-native serving with eligibility-gated continuous batching, LCP prompt reuse, structured output, speculative decoding, and sudoless silicon-bottleneck observability | Cross-platform local model workflow | Broad desktop and developer model workflows | Continuous batching, paged hot/SSD caching with prefix sharing, multi-token prediction, and distributed serving | Broad native-Swift local AI surfaces, with prompt and KV-cache reuse on the chat completions path as of v2.3.0 | SSD expert streaming for large MoE inference |
|---|