{"rewrite":{"id":"r_125ec1314625162823e9be6a","clusterId":"c_046fb31810249507d8c4186e","slug":"deepgrove-releases-maple-preview-a-small-ai-model-that-runs-on-iphone","model":"deepseek-v4-flash:free","headline":"DeepGrove Releases Maple-Preview, a Small AI Model That Runs on iPhone","summary":"DeepGrove announced Maple-Preview, a small AI model that runs on iPhone. It is a mixture-of-experts model with 20.2 billion total parameters and 1.49 billion active parameters. DeepGrove says it processes 13 times faster than Bonsai 27B on iPhone. The model is open and available under the MIT License.","whyItMatters":"Maple-Preview shows a design-time ternary approach that keeps inference speed and memory use low without quantizing an existing model, which DeepGrove argues is the wrong path.","webCardHtml":"\u003cp\u003eDeepGrove has released Maple-Preview, a small AI model built to run on iPhone. The company says the model is 13 times faster than Bonsai 27B on the same device, while using only 7.69 GB of memory when handling a 131,000-token context.\u003c/p\u003e\u003cp\u003eMaple-Preview is a mixture-of-experts model with 20.2 billion total parameters and 1.49 billion active parameters. Instead of quantizing an existing model, DeepGrove designed it to run ternary computations from the start, arguing that lowering precision after training limits both performance and efficiency.\u003c/p\u003e\u003cp\u003eThe model is open and distributed under the MIT License. DeepGrove also plans to build a system that tunes the model to individual users based on conversation content.\u003c/p\u003e","blueskyPost":"DeepGrove released Maple-Preview, an open AI model that runs on iPhone. It is 13x faster than Bonsai 27B and uses 7.69 GB for a 131K context. MIT License.","twitterPost":"DeepGrove released Maple-Preview, an open AI model that runs on iPhone. It is 13x faster than Bonsai 27B and uses 7.69 GB for a 131K context. MIT License.","threadsPost":null,"newsletterBlurb":"DeepGrove released Maple-Preview, a small AI model that runs on iPhone and is 13 times faster than Bonsai 27B. The open model uses a ternary design and is available under the MIT License.","attributionJson":"[{\"source\":\"GIGAZINE\",\"url\":\"https://gigazine.net/news/20260805-maple-preview/\",\"title\":\"iPhoneで動作しBonsai 27Bと同等性能で13倍高速なAI「Maple-Preview」が登場、ローカルAIがまた1歩前進\"}]","lintFlagsJson":null,"lintHits":0,"costUsd":0,"inputTokens":5437,"outputTokens":574,"status":"published","repairAttempts":0,"nextRepairAt":null,"factsAttemptedAt":1786380854,"createdAt":"2026-08-10T16:49:38.000Z","publishedAt":"2026-08-10T16:51:44.000Z","updatedAt":"2026-08-10T16:49:38.000Z"},"cluster":{"id":"c_046fb31810249507d8c4186e","canonicalTitle":"iPhoneで動作しBonsai 27Bと同等性能で13倍高速なAI「Maple-Preview」が登場、ローカルAIがまた1歩前進","representativeArticleId":"a_63f059ad9e2c4f2ea48463be","sourceCount":1,"writtenSourceCount":1,"writeAttempts":0,"isSolo":true,"entitiesJson":"{\"anime_titles\":[],\"manga_titles\":[],\"work_titles\":[\"Maple-Preview\",\"Bonsai 27B\"],\"studios\":[\"DeepGrove\"],\"people\":[],\"type\":\"announcement\",\"domain\":\"other\",\"is_roundup\":false}","contentType":"news","status":"published","firstSeenAt":"2026-08-05T14:00:00.000Z","lastSeenAt":"2026-08-05T14:00:00.000Z","updatedAt":"2026-08-10T16:51:44.000Z"},"attribution":[{"source":"GIGAZINE","url":"https://gigazine.net/news/20260805-maple-preview/","title":"iPhoneで動作しBonsai 27Bと同等性能で13倍高速なAI「Maple-Preview」が登場、ローカルAIがまた1歩前進"}],"entities":{"anime_titles":[],"manga_titles":[],"work_titles":["Maple-Preview","Bonsai 27B"],"studios":["DeepGrove"],"people":[],"type":"announcement","domain":"other","is_roundup":false},"keyFacts":null}