@AlexFinn I tested 31B and 26B tonight (various quants, including unsloth variants) on my RTX 5090 (32GB) and I experienced lots of performance and tool issues with OpenClaw. Hard pass. Sticking with Qwen3.5-27B which is fast and stable.
@LocallyAIApp Great to have consistency between all the *OS versions! Is there a feature coming soon to search online for current info? If I ask the Foundation or LFM2 8B what the current iPhone is, they say the iPhone 15. Both knowledge cut off in 2023/2024. This is all I need to use it more.