GLM-5.2 is now the open-weight model you benchmark your coding agent against
Z.ai's GLM-5.2 tops mid-July open-weight leaderboards with MIT license, 1M context, and ~91 GPQA — here's what that means for shipping a coding agent.
Posts tagged localllama
The "localllama" topic dives into the exciting and challenging world of self-hosted and locally deployable large language models. This field is paramount as users prioritize privacy, control, and cost-effectiveness over cloud-centric AI solutions. We explore the practicalities, hardware demands, and performance benchmarks that define this space. For example, discussions around "Kimi K3 is 2.8T params and 1.4TB of RAM — can you actually serve it?" highlight the significant infrastructure needed. We also scrutinize new releases, exemplified by "Mistral's next open-weight MoE hits early access — and the specs I actually care about are missing," and assess real-world capabilities, as seen in "MiniMax M3 self-hosted: holds up on refactors, chokes on long context." This coverage offers vital insights for anyone navigating the burgeoning landscape of accessible, local AI.
Mistral's next open-weight MoE hits early access — and the specs I actually care about are missing
Mistral's new open-weight MoE hits July early access, but the specs that matter for local use — params, VRAM, quant support — are still undisclosed.
MiniMax M3 self-hosted: holds up on refactors, chokes on long context
Ran MiniMax M3 locally against Qwen3-Coder and GLM-5.1. Strong on scoped refactors, weak on long context. A solid second model, not a daily driver.