ModelsOfficial

Thinking Machines 发布 Inkling 2 Mini:26B 循环 MoE 可在 32GB Mac 上运行

Meet Thinking Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac

Why we picked it

Thinking Machines 把编程模型塞进了消费级 Mac,3.4B 激活参数跑出 79.2% 的 SWE-bench Verified。想在本地跑代码 agent 不再需要租 GPU,写代码的开发者可以先下载试试和云端模型的差距。

Thinking Machines 发布了 Inkling 2 Mini,一个总参数量 26.4B、激活参数仅 3.4B 的循环式 MoE transformer。它在 SWE-bench Verified 上拿到 79.2% 的成绩,而这个体量的模型只需一台 32GB 内存的 Mac 或一块 DGX Spark 就能本地运行。

Image · marktechpost
Body

技术上它采用循环(looped)架构:通过重复调用同一组 MoE 层来换取推理深度,总参数大但每步激活的参数少,所以显存和算力需求远低于同规模的稠密模型。3.4B 的激活参数量让消费级硬件跑编程模型成为可能,不再依赖数据中心 GPU。

对开发者来说,这类小激活参数模型直接瞄准本地部署场景,与需要云端推理的主流编程模型形成差异。79.2% 的 SWE-bench Verified 分数说明在真实软件工程任务上它并非玩具,适合想在本地跑 agent 做代码工作的团队评估。

Original · marktechpost

Meet Thinking Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac

Thinking Machines released Inkling 2 Mini, a 26.4B-parameter looped MoE transformer with 3.4B active parameters. It runs on a 32GB Mac or one DGX Spark and scores 79.2% on SWE-bench Verified. The post Meet Thinking Machines Inkling 2 Mini: 26B Looped MoE Runs on a 32GB Mac appeared first on MarkTechPost .

Industry Coordinates & Source Verification

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, Thinking Machines, MoE, SWE-bench.

Primary source verification: published directly by marktechpost with first-party authority.