27B Model Compressed to Run on iPhone
xiaohu · x · 2026-07-15
PrismML applied model compression to Qwen3.6-27B, shrinking the original 54GB model down to 3.9–5.9GB to enable local execution on smartphones.
They compared the compressed versions against the original across 15 tests under an "intense thinking mode":
- Ternary version (5.9GB, desktop): Retains about 95% of original capabilities
- 1-bit version (3.9GB, mobile): Retains roughly 90% of capabilities
Regarding performance, the author claims speeds up to 163 token/s on an RTX 5090 and 87 token/s on an Apple M5 Max.
Related event: Bonsai 27B: The first 27B model that runs on phones(15 posts)→
More from Infra
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11
- LLM Serving Metrics Thread: Why TPOT and Uptime Make or Break User Experience — abhijithneil · 2026-09-11
- PlanetScale launches sharded Postgres: 768 servers acting as one, 1PB scale — dhruv2038 · 2026-09-11