OpenRouter measures quantized Qwen3.8 27B endpoints and finds no accuracy gap
Overview
OpenRouter says six endpoints serving Qwen3.8 27B scored 98.3% to 99.2% on 240 tasks on October 20, 2026, with no accuracy difference between fp16, fp8 and fp4.
The company's blog post also covers pricing across 1,455 provider endpoints. Lower precision was usually cheaper: 4-bit was cheaper than 8-bit for 26 of 33 models, while 8-bit cost more than 16-bit for 7 of 24 models served at both.
One fp4 endpoint scored lowest on 40 multiplication tasks. OpenRouter says that gap is too small to rule out chance.
Written by AI from the articles below · updated Nov 10, 8:11 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- OpenRouter BlogOfficialOpenRouter finds no accuracy gap between fp8, fp4 and fp16 for Qwen3.8 27B
AIOpenRouter tested six endpoints serving Qwen3.8 27B on October 20, 2026, and found no accuracy difference between fp16, fp8 and fp4, with all scoring 98.3% to 99.2% on 240 tasks. Lower precision was usually cheaper, with 4-bit cheaper than 8-bit for 26 of 33 models, though 8-bit was more expensive than 16-bit for 7 of 24 models served at both. One fp4 endpoint scored lowest on 40 multiplication tasks, a gap OpenRouter said was too small to rule out chance.
Heat Trend
- Current comparable heat
- 7
- Peak comparable heat
- 9Nov 10, 9:00 PM ET
- Change in past 24 hours
- –
The trend compares only continuously observed participants, so its scope may be smaller than the current heat total. Hover or tap for hourly heat; use arrow keys with a keyboard.