Model benchmarks

qwen/qwen3.5-9b local LLM performance

As of August 2026, qwen/qwen3.5-9b runs at up to 72.6 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).

LM Studio

Model size

9B

Peak speed

72.6 tok/s

Average speed

67.7 tok/s

Min memory

4.4 GB

Max context

8,192 tokens

Best quality

78.6

Benchmark runs

2

GPUs tested

2

Performance by hardware and tool

Every hardware/tool/quantization combination qwen/qwen3.5-9b has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 4070 TiLM Studio72.6 tok/s72.6 tok/s4.4 GB8,192 tokens57.61
NVIDIA GeForce RTX 4070 Ti SUPERLM Studio62.7 tok/s62.7 tok/s9.7 GB7,457 tokens68.91

Frequently asked questions

How fast is qwen/qwen3.5-9b for local inference?
Across 2 community benchmark runs, qwen/qwen3.5-9b reaches up to 72.6 tok/s and averages 67.7 tok/s, with the fastest results on NVIDIA GeForce RTX 4070 Ti.
How much memory does qwen/qwen3.5-9b need?
The leanest observed configuration used about 4.4 GB of memory.
Which tools have been used to run qwen/qwen3.5-9b?
Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.