Running Qwen3.6-35B-A3B-Q5_K_M with 229K context on my 5090, powering Hermes. I’m impressed how well it runs long running goals across a small KMP code base with various features. Need to play with this much more, but first look and feel is surprisingly capable.
@sudoingX I have a 5090 to use, what would you recommend for 32g? My initial setup; GLM-4.7 Flash Q4_K_M + Q4 KV + 128K but haven’t tested much yet. Would love to hear your recommendations on what models/configs to try
@pathofexile The macOS clients broke from one of the hotfixes. Failed to load instance when trying to enter from character selection and afterwards login breaks as well
@donnfelker I always do the same but never actively thought about a sophisticated reason like that, just feels right to not click the sponsored links 😅
I just ran across the article https://t.co/4SPjbBkEpU which I hoped would give good perspective on platforms and VR, but sadly has some major mistakes that make it very misleading. Let's look at those. (1/19)