Ignoring benchmarks, how do the newest local models (gemma 4 31B, 26BA4B, Qwen 3.6) “feel” to you? What do you think they compare to?

I use local ai mainly for creative writing, and benchmarks are a bit iffy on that I feel like. I’d like to compare Gemma mainly to Gemini as I like their writing the best, I do know that qwen 3.6 is amazing but mostly fo