LocalLLaMA

gemma4 vs qwen3.5 122A10 real usages

RedHatAI/gemma-4-31B-it-FP8-block vs Sehyo/Qwen3.5-122B-A10B-NVFP4 It's different quant but both are using about 90GB vram. I prefer gemma4 for financial summary. The output is concise. It also properly explaining 'resort facility' wh…