Ripple-65B Preview hits 20.5 tok/s for code but only 7.6 for prose on a single DGX Spark
DZhang50 · x · 2026-11-06
Joey ran Ricursive AI's Ripple-65B Preview on a single DGX Spark (NVFP4 weights + Ricursive's BF16 DFlash2 draft, 128K context), hooked into the Hermes agent and DeepSeek Harness with streaming and tool calls working. Measured end-to-end rates across three runs: code 20.5 tok/s, edits 26.6 tok/s, prose just 7.6 tok/s.
DFlash2 co-author DZhang50 explained the gap: their drafter was trained mostly on multi-turn code to speed up evals and RL, so slow prose speculation is expected. A vLLM backport and shorter draft blocks helped some prompts and hurt others; no clear speedup yet, so the original SGLang setup stays.
More from Infra
- Veteran investor: memory stocks aren't cheap even at P/E of 3 as LTAs are meaningless — bookwormengr · 2026-11-06
- Enterprise AI's future is local: tactical workloads favor Dell, HPE, Lenovo, Supermicro over frontier labs — DavidLinthicum · 2026-11-06
- Texas Governor-Elect: Approved Data Centers Can Be Completed, New Projects to Pay Taxes — GavinSBaker · 2026-11-06
- Marin's open 535B-A23B MoE finishes main pretraining on 18T tokens, checkpoints soon — dlwh · 2026-11-06
- Buoyant Open-Sources Kinference, an Inference Control Plane for vLLM on Kubernetes — vllm_project · 2026-11-06
- vLLM ranks #2 most-engaged open-source repo in GitHub Octoverse, #8 by contributor growth — vllm_project · 2026-11-06