syv-ai/qwen38-27b-rtx3090
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
Health Breakdown
Should you contribute to syv-ai/qwen38-27b-rtx3090?
syv-ai/qwen38-27b-rtx3090 has a FoundDev health score of 89/100, which puts it in the active-and-maintained tier. The maintainer team is shipping recently, issues are being closed, and a PR you open this week has a realistic chance of being reviewed.
Last push was 0 days ago — that signals an actively maintained project. New issues are likely to get a maintainer response within days. The project is written primarily in Python, so prior Python experience will shorten ramp-up.
Licensed under Apache-2.0, a standard OSI-approved license — safe to contribute to under normal employer IP policies.
Community
