Saya mencoba menggunakan Qwen 3.8 27B
俺のくそ雑魚スペックのPCでこれ一発で出せるんだ 一昔前のクラウドAIだったら相当驚いてたぞ
Qwen 3.8 Max telah hadir — tetapi apakah ia benar-benar memberikan semua itu?
👀 No vídeo de hoje, reagi ao benchmark da SWEN e comentei o desempenho do novo modelo, comparando os resultados e levantando uma questão importante: até onde os benchmarks realmente mostram a capacidade de
Dua agen Qwen 3.8 27B, seorang desainer dan pembuat kode, bolak-balik lebih dari 50 kali pada satu desktop sebelum proyek selesai.
128GB of memory, nothing metered. That round count is what got me. You'd never let a loop run that long against a paid API. You'd cap it at
Menguji Qwen 3.8 Flash Next Q4 pada DGX Spark: Hari ke-34.
Asked it to create Pac-Man, and after 24 minutes, it failed. DeepSeek is the only model that's come close. Interesting to see how future versions progress!
Sudah sampai ujung arsip
Semua qwen38