--- format: "aidr-story-markdown/v1" id: "f17be9431f42dabd6098d94f772cdb508e5b40589d8a76b8154ddfdf648ed98f" canonical_url: "https://aidr.today/f17be943?lang=en" title: "OpenBMB MiniCPM-V 4.7 Hits 17 Tokens Per Second on Mobile Hardware" lang: "en" requested_lang: "en" available_langs: ["en","vi"] translation_fallback: null fallback_fields: [] published_at: "2026-10-12T21:03:56.000Z" category: "Open Source" topics: ["inference","open-source"] source_urls: ["https://marketbrief.now/ai/openbmb-minicpm-v-47-hits-17-tokens-per-second-on-mobile-hardware-0ead117a","https://huggingnews.com/ai/openbmb-minicpm-v-47-hits-17-tokens-per-second-on-mobile-hardware-0ead117a"] summary: "OpenBMB has released the MiniCPM-V 4.7 multimodal AI series, featuring 35B-A3B and 1B models that utilize a Lookahead Expert Scheduling technique. The 35B-A3B variant achieves 17 tokens per second on an iPad Pro M4 and 13 tokens per second on an iPhone 17 Pro Max while using 3GB of memory. The models are available for download on GitHub and Hugging Face." --- # OpenBMB MiniCPM\-V 4\.7 Hits 17 Tokens Per Second on Mobile Hardware > [Open the canonical story]() **Published:** 2026-10-12T21:03:56.000Z **Category:** Open Source **Topics:** inference, open\-source ## Summary OpenBMB has released the MiniCPM\-V 4\.7 multimodal AI series, featuring 35B\-A3B and 1B models that utilize a Lookahead Expert Scheduling technique\. The 35B\-A3B variant achieves 17 tokens per second on an iPad Pro M4 and 13 tokens per second on an iPhone 17 Pro Max while using 3GB of memory\. The models are available for download on GitHub and Hugging Face\. ## Sources - [Story source]() - [Story source]()