Qwen3.8-27B Benchmarked on AMD R9700: Up to 227 tok/s
samsja19 · x · 2026-08-26
Benchmarks show that Qwen3.8-27B, running the Unsloth UD-IQ4XS quantization on an AMD R9700 GPU, achieves up to 227 tokens per second. Compared to a Q80 reference, the quantized version maintains 8-bit-class performance with a mean KL divergence of 0.018 and 94% top-1 agreement.
More from Models
- First test of MiniMax H3 video generation: 8s video took 1 hour on rented A100 — Winter_Assignment_78 · 2026-08-27
- Anthropic reportedly releasing Fable 5.1 model soon — mark_k · 2026-08-27
- Observers doubt Simile/Aaru claims over lack of datasets and peer review — daveholtz · 2026-08-27
- Navigator n2 released: A frontier 27B computer-use model — DhruvBatra_ · 2026-08-27
- Alibaba Releases FP8 Quantized Qwen3.8-Flash-Next Model — Qwen · 2026-08-27
- Qwen 3.8 27b coding performance shocks community, rivaling GPT 5.5 on consumer hardware — GrokiniGPT · 2026-08-27