thc1006 / qwen3.6-speculative-decoding-rtx3090View on GitHub
First public benchmark of llama.cpp speculative decoding on Qwen3.6-35B-A3B with a single RTX 3090 (post PR #19493 merge, 2026-04-19). 19 configurations covering ngram-cache, ngram-mod, and classic draft with vocab-matched Qwen3.5-0.8B. Finding: no variant achieves net speedup on Ampere + A3B MoE. Raw JSON, plots, full reproducibility.
31May 16, 2026Updated last month

Alternatives and similar repositories for qwen3.6-speculative-decoding-rtx3090

Users that are interested in qwen3.6-speculative-decoding-rtx3090 are comparing it to the libraries listed below. We may earn a commission when you buy through links labeled 'Ad' on this page.

Sorting:

Are these results useful?