Model benchmarks
local-qwen local LLM performance
As of September 2026, local-qwen runs at up to 60.5 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
60.5 tok/s
Average speed
60.2 tok/s
Avg PP
676.7 tok/s
Min memory
3.4 GB
Max context
262.144 tokens
Avg output / run
20.039 tokens
Avg runtime / run
5m 22s
Avg quality
85.9
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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 85.7 | 85.9 | 86.0 |
| Agent Workflow | 90.5 | 91.4 | 92.3 |
| Code Generation | 84.6 | 84.8 | 85.0 |
| Role Play & Narrative | 80.3 | 84.7 | 89.1 |
| Research & Analysis | 79.9 | 82.5 | 85.1 |
Performance by hardware and tool
Every hardware/tool/quantization combination local-qwen has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 7900 XTX | llama.cpp | — | 60.5 tok/s | 60.2 tok/s | 3.4 GB | 262.144 tokens | 85.9 | 2 |
Benchmark runs
All 2 local-qwen runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is local-qwen good for coding?
- In our benchmarks, local-qwen scores 84.8/100 for coding, among the top 2 for coding on consumer hardware (≤24 GB VRAM). It runs at about 60.2 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 local-qwen good for agentic (tool-using) tasks?
- In our benchmarks, local-qwen scores 91.4/100 for agentic workflows. It runs at about 60.2 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
- How fast is local-qwen for local inference?
- Across 2 community benchmark runs, local-qwen reaches up to 60.5 tok/s and averages 60.2 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
- How much memory does local-qwen need?
- The leanest observed configuration used about 3.4 GB of memory.
- Which tools have been used to run local-qwen?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.