27B Open-Source Reasoning Model Runs on iPhone
The Decoder · rss · 2026-07-15
PrismML has compressed a 27B-parameter open-source reasoning model to under 4GB, making it compact enough to run locally on an iPhone.
According to the company's benchmarks, the smallest version retains roughly 90% of the original model's performance, with almost no noticeable degradation in math and coding capabilities. The article also notes that Apple is reportedly already testing this compression technology. If true, it could help Apple close the gap in on-device AI capabilities.
More from Infra
- Lam Research: weekly token usage up ~250x in 18 months — Beth_Kindig · 2026-10-27
- Encrypted GPT-2 speedrun challenges FHE inference: best is 3.031 s/token, goal sub-1s — tensorqt · 2026-10-27
- Open-source Latch compiles PyTorch into training megakernels, 2.75x faster on H100 — sguada · 2026-10-27
- NVIDIA Claims HGX Rubin NVL8 Delivers 10x More Tokens per Megawatt Than B200 — thursdai_pod · 2026-10-27
- Together lets multiple LoRA post-training experiments share the same GPU capacity — togethercompute · 2026-10-27
- Commoditization dilemma: either human labor or GPUs become the commodity — _AustinCalvert_ · 2026-10-27