User asks for open source LLMs to run Hermes Agent locally

A Reddit user with a 4090 and 5060ti setup is seeking recommendations for open source large language models that can effectively run the Hermes Agent framework locally. The user aims to replace their current Opencode workflow, which relies on the paid MiMoV2 pro model, to avoid daily costs.

• The user has a hardware configuration consisting of two GPUs (4090 and 5060ti) and 64GB of RAM.
• They previously used Opencode with MiMoV2 pro but find the $20 daily expense unsustainable.
• They have heard reports that using Qwen3.6-27b with Hermes causes looping issues.
• The goal is to find an open source model that runs at decent tokens per second and meaningfully completes tasks.