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).
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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 54.0 | 62.4 | 76.5 |
| Agent Workflow | 72.7 | 78.3 | 85.4 |
| Code Generation | 7.5 | 29.8 | 68.3 |
| Role Play & Narrative | 77.4 | 83.9 | 88.6 |
| Research & Analysis | 42.9 | 57.5 | 75.1 |
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.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon Graphics | llama.cpp | — | 97.5 tok/s | 96.7 tok/s | 15.5 GB | 98,304 tokens | 62.4 | 7 |
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.