StayLameBro's Backburner Uses iPhone Memory to Push Apple Silicon Macs Past 64K-Token AI Context
Summary
StayLameBro releases the pre-release Backburner, which uses an iPhone 17 Pro Max over 10 Gb/s USB-C as memory for an Apple Silicon Mac, pushing local Qwen3.8-27B AI context beyond 64,000 tokens to an estimated 196,000-229,000 tokens.
Key Points
- StayLameBro releases Backburner, a pre-release that uses an iPhone over 10 Gb/s USB-C to help an Apple Silicon Mac run Qwen3.8-27B locally, accelerating prompt prefill and extending context.
- On a 24 GB M4 Pro MacBook with an iPhone 17 Pro Max, Backburner reports 29-44% faster 2,000-token prompt reads at 16k-48k context, reaching 157 tokens per second at 16k.
- The system keeps 8-bit context beyond the Mac's 64k-token limit by moving old KV pages to the phone, sizing total capacity at 196k-229k tokens on an iPhone 17 Pro Max and testing to 128k at 8-bit.