KVarN
KVarN is a native vLLM KV-cache quantization backend for your agents: 3-5x more context, throughput above FP16, and FP16-level accuracy. Calibration-free, one flag.
huawei-csl/KVarN
View on GitHub ↗KVarN is a native vLLM KV-cache quantization backend for your agents: 3-5x more context, throughput above FP16, and FP16-level accuracy. Calibration-free, one flag.
KVarN grew +1.6% in the last 30 days, ranks 95th of 305 in Inference Optimization, last commit 126 days ago.
Star trajectory
116 days tracked2026-07-02513 stars today2026-10-26
Momentum
+1.6% in 30d, above the Inference Optimization median of +1.6%.
Maintenance
Last commit 126 days ago, staler than the median of 15d.
How it compares in Inference Optimization
KVarN
513
stars
+1.6% 30d
commit 126 days ago
Inference Optimization median
254
stars
+1.6% 30d
commit ~15d ago
Leader · supertonic
13.8k
stars
-0.2% 30d
commit 47 days ago
Ranked 95th of 305 by stars.
- Language
- Python
- License
- Apache License 2.0
- Contributors
- 3
- Forks
- 37
- Watchers
- 5
- Open issues
- 8
- Releases
- 0
- Created
- 2026-05-29
- Stars this week
- +3
You've seen where KVarN stands today.
Get the moves before everyone else: the 5 Inference Optimization repos climbing fastest, every Tuesday.
Free weekly digest. Double opt-in, unsubscribe anytime.