Qwen3.8-27B PrismaAqua benchmark: Near-BF16 quality
offgridai · reddit · 2026-08-18
A user benchmarked the 5.5-bit PrismaAqua quantization of Qwen 3.8 27B on an RTX 5090 using vLLM, focusing on tool use, business reasoning, and investment tasks. Results show the quantized model closely matches BF16 performance in a custom test suite, achieving 95% accuracy in tool use. The user found that setting reasoning effort to "medium" outperformed "xhigh" by avoiding token limit timeouts while maintaining quality. Specific vLLM configuration parameters and throughput metrics are provided.
More from Models
- Prediction: frontier AI revenue growth could slow even as AI usage keeps rising — yi111 · 2026-10-16
- Meta teased 816B-param open-weight Muse Spark Open with 1M context, says meme thread — alexandr_wang · 2026-10-16
- Nolla Health Post-Trains Nemotron 3.5 VL: Visual Diagnosis Jumps from 35% to 66.8% — LoganGrasby · 2026-10-16
- Dev argues coding models now beat nearly all engineers in every branch of software — zack_overflow · 2026-10-16
- Prediction: Chinese frontier models could match GPT-6 by March 2027 — teortaxesTex · 2026-10-16
- Liquid AI's d1-ultra, #1 on Artificial Analysis Decision Index, lands on Vercel — JosephJacks_ · 2026-10-16