Is qwen3.5:9b good for coding?

In our benchmarks, qwen3.5:9b scores 50.7/100 for coding. It runs at about 54.5 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).

Is qwen3.5:9b good for agentic (tool-using) tasks?

In our benchmarks, qwen3.5:9b scores 71.1/100 for agentic workflows. It runs at about 54.5 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).

How fast is qwen3.5:9b for local inference?

Across 17 community benchmark runs, qwen3.5:9b reaches up to 101.4 tok/s and averages 54.5 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.

How much memory does qwen3.5:9b need?

The leanest observed configuration used about 5.3 GB of memory (quantizations tested: Q4_K_M).

Which tools have been used to run qwen3.5:9b?

Benchmarks were submitted using LM Studio, Ollama. Results are community-contributed and updated as new runs arrive.

Model benchmarks

qwen3.5:9b local LLM performance

As of October 2026, qwen3.5:9b runs at up to 101.4 tok/s for local inference (best of 17 community benchmark runs across 8 GPUs).

LM StudioOllamaQ4_K_M

ShareReddit X

Model size

9.7B

Peak speed

101.4 tok/s

Average speed

54.5 tok/s

Avg PP

1536.9 tok/s

Min memory

5.3 GB

Max context

65,536 tokens

Avg output / run

18,264 tokens

Avg runtime / run

4m 25s

Avg quality

66.2

Benchmark runs

GPUs tested

Quality by task

Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 17 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

Task	P5 (low)	Avg	P95 (high)
Overall	58.9	66.2	73.3
Agent Workflow	53.7	71.1	80.5
Code Generation	29.2	50.7	63.1
Role Play & Narrative	63.0	76.6	84.8
Research & Analysis	49.6	66.4	78.1

Performance by hardware and tool

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

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
AMD Radeon RX 7900 XTX	LM Studio	—	101.4 tok/s	101.4 tok/s	n/a	8,192 tokens	70.0	1
AMD Radeon RX 9070 XT	Ollama	Q4_K_M	85.3 tok/s	85.3 tok/s	9.1 GB	65,536 tokens	63.1	1
NVIDIA GeForce RTX 5070	LM Studio	—	78.6 tok/s	78.5 tok/s	n/a	65,536 tokens	72.6	2
NVIDIA GeForce RTX 4070 Ti	LM Studio	—	72.6 tok/s	72.6 tok/s	n/a	16,384 tokens	68.5	1
AMD Radeon RX 7800 XT	Ollama	Q4_K_M	71.9 tok/s	69.0 tok/s	5.3 GB	65,536 tokens	63.0	4
NVIDIA GeForce RTX 4070 Ti SUPER	LM Studio	—	62.7 tok/s	62.7 tok/s	n/a	16,384 tokens	72.9	1
AMD Radeon RX 6800/6800 XT / 6900 XT	LM Studio	Q4_K_M	58.1 tok/s	58.1 tok/s	8.2 GB	32,768 tokens	67.6	1
NVIDIA GeForce RTX 5080	Ollama	Q4_K_M	24.1 tok/s	22.3 tok/s	8.2 GB	16,384 tokens	61.6	2
AMD Radeon RX 7900 XTX	Ollama	Q4_K_M	19.8 tok/s	17.1 tok/s	8.6 GB	65,536 tokens	65.8	4

Benchmark runs

All 17 qwen3.5:9b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Open run page →

Frequently asked questions

Is qwen3.5:9b good for coding?: In our benchmarks, qwen3.5:9b scores 50.7/100 for coding. It runs at about 54.5 tok/s, so if you want more speed, Nex-N2.5-mini-IQ3_XXS is faster (~91.4 tok/s) and still scores well for coding (81.6/100).
Is qwen3.5:9b good for agentic (tool-using) tasks?: In our benchmarks, qwen3.5:9b scores 71.1/100 for agentic workflows. It runs at about 54.5 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
How fast is qwen3.5:9b for local inference?: Across 17 community benchmark runs, qwen3.5:9b reaches up to 101.4 tok/s and averages 54.5 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does qwen3.5:9b need?: The leanest observed configuration used about 5.3 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run qwen3.5:9b?: Benchmarks were submitted using LM Studio, Ollama. Results are community-contributed and updated as new runs arrive.