Model benchmarks

qwen36-agent local LLM performance

As of November 2026, qwen36-agent runs at up to 97.5 tok/s for local inference (best of 7 community benchmark runs across 0 GPUs).

llama.cpp
ShareRedditX

Model size

Unknown

Peak speed

97.5 tok/s

Average speed

96.7 tok/s

Avg PP

1486.1 tok/s

Min memory

15.5 GB

Max context

98,304 tokens

Avg output / run

17,159 tokens

Avg runtime / run

3m 3s

Avg quality

62.4

Benchmark runs

7

GPUs tested

0

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 7 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)
Overall54.062.476.5
Agent Workflow72.778.385.4
Code Generation7.529.868.3
Role Play & Narrative77.483.988.6
Research & Analysis42.957.575.1

Performance by hardware and tool

Every hardware/tool/quantization combination qwen36-agent has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon Graphicsllama.cpp—97.5 tok/s96.7 tok/s15.5 GB98,304 tokens62.47

Benchmark runs

All 7 qwen36-agent runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is qwen36-agent good for agentic (tool-using) tasks?
In our benchmarks, qwen36-agent scores 78.3/100 for agentic workflows. It runs at about 96.7 tok/s, so if you want more speed, cyber-tiel-coder-35b-a3b-mtp is faster (~115.9 tok/s) and still scores well for agentic workflows (86.4/100).
How fast is qwen36-agent for local inference?
Across 7 community benchmark runs, qwen36-agent reaches up to 97.5 tok/s and averages 96.7 tok/s, with the fastest results on AMD Radeon Graphics.
How much memory does qwen36-agent need?
The leanest observed configuration used about 15.5 GB of memory.
Which tools have been used to run qwen36-agent?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.