Back to Research
Independent research

Speculative decoding on an A100 — what actually got faster

The headline is 1.39×. The finding worth reading is that the other method I tested was three times slower.

Results

Every value links to the committed artifact that produces it.

MetricValueEvidence
Aggregate latency speedup — vLLM + EAGLE-3, Qwen3-8B, A100 80 GB1.387×paper/data/phase3_results.csv
Peak on math-reasoning promptsone category; a latency speedup, not throughputup to 1.46× latencyextracted_metrics.json:147
Global throughput110.93 tok/sextracted_metrics.json:146
Study 1 (Transformers assisted decoding)a negative result, reported as "Study 1: no acceleration"0.25–0.36× — ~3× slowerpaper/data/phase3_results.csv

What didn’t work

Transformers-assisted decoding came out ~3× slower than the baseline. The README says so outright; the study reports the confound rather than hiding it.

Reported as a null, not hidden — the same honesty policy applies to every number on this site.

Running it

Live decoding needs 80 GB of VRAM; the site shows a dashboard built from the committed extracted_metrics.json plus the notebook as read-only HTML.

Stack

  • vLLM
  • EAGLE-3
  • Speculative decoding
  • Qwen3-8B
  • A100 80GB