Artificial Analysis benchmarks Qwen3.5 9B
Artificial Analysis has published a benchmark page for Qwen3.5 9B with reasoning and non-reasoning variants, reporting a 1.08-second time to first answer token for the non-reasoning model.
Independent benchmarking firm Artificial Analysis has published results for Qwen3.5 9B. The release comes in two models, a reasoning variant and a non-reasoning variant, and the page compares them on intelligence, cost, speed and latency.
The page gives one specific figure: the non-reasoning Qwen3.5 9B has a time to first answer token of 1.08 seconds, the lower of the two variants.
Under Artificial Analysis's method, each model is compared only with models in its own class. Non-reasoning models are ranked against other non-reasoning models. Reasoning models are ranked against both kinds. Open-weights models are grouped by parameter count, and the "Small" tier covers 4B to 40B parameters. Proprietary models are grouped by blended price, using a 3:1 input-to-output token ratio.
The page also lists the evaluations behind its capability indexes. These include AA-Briefcase v1.1, AA-Omniscience v1.1, HLE-Diamond, AutomationBench-AA, GDP.pdf, MLCR-AA, CritPt v2 (Hard Subset), Terminal-Bench 4.0, Terminal-Bench-Science 0.1 and FrontierCode-AA 2.0.
The material provided does not include the models' actual Intelligence Index scores, cost per task, output speed or pricing, so how Qwen3.5 9B ranks against its peers is not yet clear. It also gives no details from the developer on licensing, context length or training.