Model benchmarks

qwen3.6:latest local LLM performance

As of June 2026, qwen3.6:latest runs at up to 25.6 tok/s for local inference (best of 6 community benchmark runs across 1 GPU).

OllamaQ4_K_M

Model size

36.0B

Peak speed

25.6 tok/s

Average speed

9.1 tok/s

Min memory

22.9 GB

Max context

7,654 tokens

Best quality

92.7

Benchmark runs

6

GPUs tested

1

Performance by hardware and tool

Every hardware/tool/quantization combination qwen3.6:latest has been benchmarked on, ranked by peak token generation speed. Last updated June 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 7900 XTXOllamaQ4_K_M25.6 tok/s9.1 tok/s22.9 GB7,654 tokens78.06

Frequently asked questions

How fast is qwen3.6:latest for local inference?
Across 6 community benchmark runs, qwen3.6:latest reaches up to 25.6 tok/s and averages 9.1 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
How much memory does qwen3.6:latest need?
The leanest observed configuration used about 22.9 GB of memory (quantizations tested: Q4_K_M).
Which tools have been used to run qwen3.6:latest?
Benchmarks were submitted using Ollama. Results are community-contributed and updated as new runs arrive.