88Z AI, 20 September — MiniMax has open-sourced version v0.4.12 of MiniMax Code CLI, with the source code now available on GitHub. As a core component of the MiniMax Code client, the tool achieved a 76.7% task pass rate in the FrontierHarness Eval benchmark, with a median time of 4 minutes and 33 seconds for successful tasks, outperforming public baselines. The company states that this open-source release allows developers to inspect tool invocation and permission handling, fostering the joint creation of secure and r
At a glance
- MiniMax Code CLI v0.4.12 source code is now open on GitHub, outperforming public baselines.
- The open-source release enables developers to inspect tool calls and permissions for secure enterprise harnesses.
The update
MiniMax has officially released the open-source code for version v0.4.12 of MiniMax Code CLI, with the files now hosted on GitHub. As a critical core component of the MiniMax Code client, this command-line tool demonstrated strong performance in the recent FrontierHarness Eval benchmark. Data shows it achieved a task pass rate of 76.7%, with a median completion time of just 4 minutes and 33 seconds for successful tasks, significantly outperforming existing public baselines. The company stated that the primary goal of this open-source initiative is to provide developers with greater transparency, allowing them to closely inspect tool invocation logic and permission handling processes. By opening up the source code, MiniMax aims to collaborate with the community to build a secure and reliable enterprise-grade harness environment, thereby enhancing the trustworthiness and stability of AI coding tools in practical applications.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- MiniMax 稀宇科技Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.