ThinkyMachines 的 Inkling 2 Mini 登陆 Ollama,5.4B 参数可本地运行
ThinkyMachines 把 5.4B 的 Inkling 2 Mini 放进了 Ollama,32GB 内存 Mac 就能跑。它靠中间层循环三次来加深推理,主打编码和 computer use 这类 agent 任务。想在本地折腾小模型做编程助手的人可以直接装来试试。
ThinkyMachines 的 Inkling 2 Mini 已上架 Ollama 0.40.17 预发布版本,本地一条命令 ollama run inkling-2-mini 即可运行。模型仅 5.4B 激活参数,32GB 内存的 Mac 就能跑。
它的核心机制是循环复用中间层:每个 token 会经过中段层三次,用少量参数换取更深的推理能力。官方定位面向编程、智能体工作流(如 computer use)、带强视觉推理的设计任务以及金融场景,视频演示中它实时决策玩 2048。
对小参数模型来说,用层循环替代单纯堆参数是一条差异化路线,对想在本地跑编码和 agent 任务的开发者,这是一个无需云端 API 的选项。
Inkling 2 Mini from @ThinkyMachines 🇺🇸 is now on Ollama with 0.40.17 (available in pre-release) Small enough to run on a Mac with 32GB+ memory. Each token loops through its middle layers three times so it reasons deeper at just 5.4B active parameters. Built for coding, agentic workflows such as, computer use, design with strong visual reasoning, and finance tasks. Here’s Inkling 2 Mini playing 2048 by making decisions in real-time 💨 Run it with Ollama fully locally: ollama run inkling-2-mini Run it in Pi: ollama launch pi --model inkling-2-mini
This report is categorized under AITOP's "Models" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows and US-China foundation model benchmarks and commercialization comparison. Key entities and technical topics under continuous tracking include Inkling 2 Mini, ThinkyMachines, Ollama, reasoning models.
Source attribution: curated from ollama (@ollama).