← Leaderboard
qwen/qwen3.6-27b

Qwen3.6-27B (FP8)

vLLM 27B AgentsGeneralReasoning
Throughput
27.5t/s@ 256k
Engine
vLLM
Parameters
27B
Released
2026-04-21
Benchmarked
2026-06-27

Context ladder

Throughput at each benched context window (single measurement).

Context KV Throughput
@ 256k peak golden 27.5t/s

Golden profile

opencode-qwen27-dflash-262k

Capabilities

generaldensevllmfp8

Why we run it

Official Qwen FP8 checkpoint — reference vLLM path for the dense 27B base model.

Bench notes

golden ?/? @ 3.7 tok/s — fill~50000 — bench-agent-v2 — tool_ok=True

Benchmarked 2026-06-27
SparkBench · GB10 · single node