Model benchmarks

Qwen3.6-35B-A3B-UD-IQ3_XXS local LLM performance

As of November 2026, Qwen3.6-35B-A3B-UD-IQ3_XXS runs at up to 100.4 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

llama.cpp
ShareRedditX

Model size

35B

Peak speed

100.4 tok/s

Average speed

56.2 tok/s

Avg PP

600.8 tok/s

Min memory

13.2 GB

Max context

98,304 tokens

Avg output / run

12,485 tokens

Avg runtime / run

6m 33s

Avg quality

69.7

Benchmark runs

2

GPUs tested

1

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 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall66.169.773.3
Agent Workflow61.368.074.6
Code Generation66.469.271.9
Role Play & Narrative64.170.376.5
Research & Analysis70.471.472.5

Performance by hardware and tool

Every hardware/tool/quantization combination Qwen3.6-35B-A3B-UD-IQ3_XXS has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon Graphicsllama.cpp—100.4 tok/s100.4 tok/s16.3 GB98,304 tokens73.71
AMD Radeon 680Mllama.cpp—12.0 tok/s12.0 tok/s13.2 GB8,192 tokens65.71

Benchmark runs

All 2 Qwen3.6-35B-A3B-UD-IQ3_XXS runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is Qwen3.6-35B-A3B-UD-IQ3_XXS good for coding?
In our benchmarks, Qwen3.6-35B-A3B-UD-IQ3_XXS scores 69.2/100 for coding. It runs at about 56.2 tok/s, so if you want more speed, qwen3.8-27b-Swift1.5-Uncensored-ATX-AI3 is faster (~85.5 tok/s) and still scores well for coding (84.8/100).
Is Qwen3.6-35B-A3B-UD-IQ3_XXS good for agentic (tool-using) tasks?
In our benchmarks, Qwen3.6-35B-A3B-UD-IQ3_XXS scores 68.0/100 for agentic workflows. It runs at about 56.2 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 Qwen3.6-35B-A3B-UD-IQ3_XXS for local inference?
Across 2 community benchmark runs, Qwen3.6-35B-A3B-UD-IQ3_XXS reaches up to 100.4 tok/s and averages 56.2 tok/s, with the fastest results on AMD Radeon Graphics.
How much memory does Qwen3.6-35B-A3B-UD-IQ3_XXS need?
The leanest observed configuration used about 13.2 GB of memory.
Which tools have been used to run Qwen3.6-35B-A3B-UD-IQ3_XXS?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.