{"rewrite":{"id":"r_91de6fe718cff36366b23afd","clusterId":"c_326ad582c96433c6d62553d6","slug":"prismml-releases-bonsai-27b-a-27b-parameter-ai-model-that-runs-on-iphone-17-pro","model":"deepseek-v4-flash","headline":"PrismML Releases Bonsai 27B, a 27B-Parameter AI Model That Runs on iPhone 17 Pro","summary":"PrismML released Bonsai 27B, a compressed version of Qwen3.6-27B that runs on iPhone 17 Pro. The model comes in two variants: a 1-bit version at 3.9GB that retains about 90% of the original's performance, and a ternary version at 5.9GB that retains about 95%. It handles up to 262,000 tokens of context, supports multimodal inputs and tool calling, and is published under the Apache 2.0 license.","whyItMatters":"The release demonstrates that a 27-billion-parameter model can be compressed to fit within a phone's memory while preserving most of its original capability, making on-device inference practical for offline, private use.","webCardHtml":"\u003cp\u003eThe 1-bit variant uses only 3.9GB of memory, placing it within the app-accessible memory limit on iPhone 17 Pro. The ternary variant uses 5.9GB. On a 15-benchmark aggregate, the original Qwen3.6-27B scored 85.0; the ternary version scored 80.5, and the 1-bit version scored 76.1. PrismML reports the ternary variant maintains about 95% of the original model\u0026#39;s performance and the 1-bit variant about 90%.\u003c/p\u003e\u003cp\u003eBonsai 27B supports a context window of roughly 262,000 tokens and accepts text, screenshots, documents, and camera images as input. It can call external tools and handle agent-style workflows. Under Apache 2.0, the model weights are available on Hugging Face in MLX (Apple) and GGUF (llama.cpp) formats.\u003c/p\u003e","blueskyPost":"PrismML released Bonsai 27B, a compressed 27B-parameter AI model that runs on iPhone 17 Pro. Two variants: 1-bit (3.9GB, 90% of original performance) and ternary (5.9GB, 95%). Open source under Apache 2.0.","twitterPost":"PrismML released Bonsai 27B, a compressed 27B-parameter AI model that runs on iPhone 17 Pro. Two variants: 1-bit (3.9GB, ~90% perf) and ternary (5.9GB, ~95% perf). Apache 2.0, on Hugging Face.","threadsPost":null,"newsletterBlurb":"PrismML announced Bonsai 27B, a compressed AI model that runs on iPhone 17 Pro. The 1-bit variant uses 3.9GB and retains 90% of the original Qwen3.6-27B's performance. The model is available under Apache 2.0.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260715-bonsai-27b/\",\"title\":\"iPhoneでローカル実行可能な270億パラメータのAIモデル「Bonsai 27B」が登場、Qwen3.6-27Bをメモリ使用量3.9GBまで小型化してスマホ単体で実用的な速度で動作\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5389,"outputTokens":1602,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1784088714,"createdAt":"2026-07-15T04:07:20.000Z","publishedAt":"2026-07-15T04:07:29.000Z","updatedAt":"2026-07-15T04:07:20.000Z"},"cluster":{"id":"c_326ad582c96433c6d62553d6","canonicalTitle":"iPhoneでローカル実行可能な270億パラメータのAIモデル「Bonsai 27B」が登場、Qwen3.6-27Bをメモリ使用量3.9GBまで小型化してスマホ単体で実用的な速度で動作","representativeArticleId":"a_980beec17ee821454f6d9a5f","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[],\"studios\":[],\"people\":[],\"type\":\"news\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-07-15T03:10:00.000Z","lastSeenAt":"2026-07-15T03:10:00.000Z","updatedAt":"2026-07-15T04:07:29.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260715-bonsai-27b/","title":"iPhoneでローカル実行可能な270億パラメータのAIモデル「Bonsai 27B」が登場、Qwen3.6-27Bをメモリ使用量3.9GBまで小型化してスマホ単体で実用的な速度で動作"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":[],"studios":[],"people":[],"type":"news","domain":"other","is_roundup":false},"keyFacts":null}