Is qwen/qwen3.6-35b-a3b good for coding?

In our benchmarks, qwen/qwen3.6-35b-a3b scores 60.6/100 for coding. It runs at about 70.0 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).

Is qwen/qwen3.6-35b-a3b good for agentic (tool-using) tasks?

In our benchmarks, qwen/qwen3.6-35b-a3b scores 70.1/100 for agentic workflows. It runs at about 70.0 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~165.0 tok/s) and still scores well for agentic workflows (89.3/100).

How fast is qwen/qwen3.6-35b-a3b for local inference?

Across 12 community benchmark runs, qwen/qwen3.6-35b-a3b reaches up to 437.3 tok/s and averages 70.0 tok/s, with the fastest results on NVIDIA GeForce RTX 5090.

How much memory does qwen/qwen3.6-35b-a3b need?

The leanest observed configuration used about 14.4 GB of memory.

Which tools have been used to run qwen/qwen3.6-35b-a3b?

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

Model benchmarks

qwen/qwen3.6-35b-a3b local LLM performance

As of August 2026, qwen/qwen3.6-35b-a3b runs at up to 437.3 tok/s for local inference (best of 12 community benchmark runs across 5 GPUs).

LM StudiovLLM

ShareReddit X

Model size

35B

Peak speed

437.3 tok/s

Average speed

70.0 tok/s

Min memory

14.4 GB

Max context

131.072 tokens

Avg output / run

23.114 tokens

Avg runtime / run

6m 1s

Avg quality

73.1

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 12 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	56.9	73.1	83.6
Agent Workflow	35.8	70.1	85.9
Code Generation	28.0	60.6	79.5
Role Play & Narrative	72.1	82.7	92.2
Research & Analysis	68.0	78.8	88.6

Performance by hardware and tool

Every hardware/tool/quantization combination qwen/qwen3.6-35b-a3b has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.

Hardware	Tool	Quant	Best tok/s	Avg tok/s	Memory	Context	Quality	Runs
NVIDIA GeForce RTX 5090	vLLM	—	437.3 tok/s	437.3 tok/s	26.5 GB	65.536 tokens	84.7	1
NVIDIA GeForce RTX 5070 Ti	LM Studio	—	75.9 tok/s	75.9 tok/s	14.4 GB	121.072 tokens	70.8	1
Apple M4 Max	LM Studio	—	69.8 tok/s	69.8 tok/s	35.2 GB	120.000 tokens	72.9	1
AMD Radeon RX 7900 XTX	LM Studio	—	46.1 tok/s	31.1 tok/s	17.1 GB	65.536 tokens	68.1	6
CPU only	LM Studio	—	25.0 tok/s	24.9 tok/s	17.1 GB	131.072 tokens	78.7	2
AMD Radeon RX 7600 XT	LM Studio	—	20.8 tok/s	20.8 tok/s	21.1 GB	32.768 tokens	82.2	1

Benchmark runs

All 12 qwen/qwen3.6-35b-a3b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is qwen/qwen3.6-35b-a3b good for coding?: In our benchmarks, qwen/qwen3.6-35b-a3b scores 60.6/100 for coding. It runs at about 70.0 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
Is qwen/qwen3.6-35b-a3b good for agentic (tool-using) tasks?: In our benchmarks, qwen/qwen3.6-35b-a3b scores 70.1/100 for agentic workflows. It runs at about 70.0 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~165.0 tok/s) and still scores well for agentic workflows (89.3/100).
How fast is qwen/qwen3.6-35b-a3b for local inference?: Across 12 community benchmark runs, qwen/qwen3.6-35b-a3b reaches up to 437.3 tok/s and averages 70.0 tok/s, with the fastest results on NVIDIA GeForce RTX 5090.
How much memory does qwen/qwen3.6-35b-a3b need?: The leanest observed configuration used about 14.4 GB of memory.
Which tools have been used to run qwen/qwen3.6-35b-a3b?: Benchmarks were submitted using LM Studio, vLLM. Results are community-contributed and updated as new runs arrive.