● Covered by 1 source · 1 reportMedium impact1 neutral
Slotstream enables running 104GB Qwen3.8-Flash-Next AI model on 48GB Mac via SSD streaming
Key points
- Runs 104GB Qwen3.8-Flash-Next on Macs with 48GB RAM.
- Streams model weights from SSD, requiring ~110GB free disk space.
- Swift binary, no Python, supports Ollama and OpenAI chat APIs.
- Installation via curl script, with prebuilt binaries and source build options.
Running Large AI Models on Macs
Slotstream is a new tool that enables Apple Silicon Mac users to run the Qwen3.8-Flash-Next 125B-parameter mixture-of-experts AI model, which has a 104GB footprint, on systems with as little as 48GB of RAM. It achieves this by streaming the model's weights from the SSD, allowing the model to operate within available memory constraints.
Technical Implementation and Requirements
The tool is a single Swift binary, eliminating the need for Python dependencies. It is compatible with macOS 14+ and requires approximately 110GB of free disk space for the model weights. Slotstream supports the Ollama and OpenAI chat APIs, ensuring compatibility with existing AI development tools. Users can check their system's compatibility and estimated performance using the `slotstream doctor` command.
Installation and Model Management
Installation is performed via a curl script that places a prebuilt binary in the user's PATH. The tool also supports building from source using Command Line Tools. Model weights, totaling 103.8GB across 24 files, are downloaded on first use or via the `slotstream pull` command. The download process includes SHA256 hash verification and resume capabilities for interrupted transfers, ensuring data integrity.
Usage and Performance
Users can interact with the model directly via the `slotstream run` command for single prompts or use `slotstream serve` to expose an API endpoint on port 11434, compatible with Ollama clients and OpenAI SDKs. The model runs at approximately 12 tokens per second on a 48GB Mac. Download speeds for the model weights are capped by Hugging Face at 36–57 MB/s, meaning a full download can take several hours depending on internet connection speed.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
The daily brief
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Today's brief
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's brief
New every morning, and the back catalogue is archived by date.
Primary sources
GitHub cmc_internal/api GitHub github/collect GitHub carloslfu/slotstream GitHub _private/browser GitHub get-started/accessibility GitHub open-source/sponsorsReporting from
Slotstream, a new Swift binary, allows users to run the 104GB Qwen3.8-Flash-Next mixture-of-experts AI model on Apple Silicon Macs with less RAM by streaming model weights from SSD. This tool provides a method for Mac users to access large language models locally that would otherwise exceed their system's memory capacity.