Qwen3.8-27B on a single RTX 3090 with vLLM: ~1,000 tok/s at 64 concurrent (int8 tensor-core GEMMs, fp16 DeltaNet state), ~114 tok/s single-user at default sampling / ~124 greedy (MTP drafts, own-output draft vocab, calibrated int4 lm_head, split-KV verify attention), 150k-262k context; patches, requant scripts, benchmarks
syv-ai/qwen38-27b-rtx3090 is accelerating — star growth is speeding up by +15.4/day (95th percentile in the tracked cohort), on top of +34.4 stars/day. Forks are climbing too, at +2.6/day.
Elevated breakout odds over the next 14 days, led by star acceleration (95th pctl) and star velocity (88th pctl). Confidence is high given the available history.
Reconstructed from 64 snapshots — the time-series GitHub’s API doesn’t expose.
Live score, refreshed every 6 hours. Links back to this page.
[](https://repoclimb.com/repo/syv-ai/qwen38-27b-rtx3090)Transparent heuristic · logged for model training
Percentile rank within the tracked cohort. The score self-calibrates — it answers “accelerating vs. everything else,” not raw size.