Benchmark - 4x 5060 Ti (64GB VRAM) (P2P) - Qwen3.6 27B (INT8 /w bf16 kv cache) @ 8 concurrency with SGLang. SGLang seems to handle higher concurrency better with this setup
Benchmark - 4x 5060 Ti (64GB VRAM) (P2P) - Qwen3.6 27B (INT8 /w bf16 kv cache) @ 8 concurrency with SGLang. SGLang seems to handle higher concurrency better with this setup — reported by reddit.com, aggregated and ranked by ClawDigest.