Free 55% Decoding Speedup: Raising llama.cpp --adapt-every from 1 to 6 on Qwen Flash
Bzucko · reddit · 2026-10-17
The author found that raising llama.cpp's --adapt-every from the calibrated default of 1 to 6 substantially boosts decoding speed when running Qwen 3.8 Flash Next locally (GGUF quants):
- UD-Q4KXL: +19%, +12%, +23%, and +55% across four Slovak-language chat prompts (the longest, a 20K-token source-code description, went from 27.8 to 43.1 t/s)
- UD-IQ4XS: +9%, +24%, +24%, +16%
- Test rig: Intel [email protected], RTX 4070 Ti Super, 64GB DDR4, Arch Linux TTY
Gains are largest on long outputs — a zero-cost local inference tuning trick.
More from Infra
- Reliance welcomes Starlink competition in India: 11,000+ satellites in orbit vs Jio's 2027 test launch — elonmusk · 2026-10-17
- Microsoft completes pre-training of MAI-Base-2: 2.8T-param MoE, 4-12x more FLOPs-efficient than DeepSeek-V4-Pro-Base — NandoDF · 2026-10-17
- Apple research: a 7B model with 8 retries matches a 32B model using 4x less memory — CatAstro_Piyush · 2026-10-17
- Can a ternarized DeepSeek run on a single RTX 3090? Open problem: agent accuracy — casper_hansen_ · 2026-10-17
- Running LLMs on a Commodore 64, a 486 DOS PC and a Nintendo Wii — Awacer · 2026-10-17
- Buy vs rent vs API for inference: it all hinges on your sustained GPU utilization — Dizzy_Lingonberry947 · 2026-10-17