vLLM 0.12 on 4x Arc Pro B60 (TP=4): Qwen3-VL-30B chat 1K->1K TTFT (concurrency 16)
2100 ms
Fields
id
benchmarks/embeddedllm-b60x4-vllm-qwen3vl30b-chat-ttftMetric
ttft_ms
Value
2,100
Unit
ms
Model
Qwen/Qwen3-VL-30B-A3B-Instruct
Methodology
THIRD-PARTY MEASURED (Embedded LLM, Feb 14 2026). Sparkle 4x Intel Arc Pro B60 workstation, TP=4, 24GB/456GB-s per GPU. Qwen/Qwen3-VL-30B-A3B-Instruct. vLLM 0.12 vs Intel LLM-Scaler 1.2. Concurrency 16/32/64, 1K->1K and 1K->2K scenarios. Chat scenario 1K->1K at concurrency 16. TTFT = request->first-token.
Measured by
third_party
Reproducible
yes
notes
vLLM 0.12 standard build, lower TTFT than LLM-Scaler (4.2s) for chat. Per-request TTFT at c=16.
Notes
vLLM 0.12 standard build, lower TTFT than LLM-Scaler (4.2s) for chat. Per-request TTFT at c=16.