A framework for efficient model inference with omni-modality models
vllm-project/vllm-omni is accelerating — star growth is speeding up by +14.6/day (97th percentile in the tracked cohort), on top of +28.4 stars/day. A recent release is helping fuel the run.
Reconstructed from 209 snapshots — the time-series GitHub’s API doesn’t expose.
~13 forks/day, 3.2× the trailing baseline.
Release count rose to 28.
~24 forks/day, 5.4× the trailing baseline.
Live score, refreshed every 6 hours. Links back to this page.
[](https://breakwave.vercel.app/repo/vllm-project/vllm-omni)Elevated breakout odds over the next 14 days, led by star acceleration (97th pctl) and star velocity (86th pctl). Confidence is high given the available history.
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.
~23 forks/day, 4.6× the trailing baseline.
~15 forks/day, 3.2× the trailing baseline.
Release count rose to 27.
~21 forks/day, 3.1× the trailing baseline.
~21 forks/day, 4.6× the trailing baseline.
Release count rose to 26.
Release count rose to 25.
~35 forks/day, 9.9× the trailing baseline.
~14 forks/day, 3.4× the trailing baseline.