Qwen 3.8 27B running on a 3090 at Q5 just beat gemini 3.7 flash on a water surface simulation.
Same prompt, one shot, not full precision. One of the best water simulations seen across every local model tested.
Yeah I know it’s August but the $CKB blocks don’t wait.
ThreeJS and AI gonna make the festive season beautiful this year. GLM5.2 vs Grok 4.6 vs Gemini 3.7 vs Qwen 3.8. For the first time Gemini comes out on top here. GLM solid overall in 2nd place. Grok surprisingly lakes the L.
Benchmarking Qwen 3.8
Muse Glimmer CrackNzT Will share benchmarks soon I promise
Qwen 3.8 max just beat GPT-5.6 sol max at actually using a computer.
But the benchmark isn't even the most interesting part. The numbers: → 2.4 trillion parameter Mixture-of-Experts model → Only ~95B parameters activate per token → 1M token context for huge SOPs,
You have reached the end of the archive
All of qwen38