Model benchmarks
Qwen-MidTemp local LLM performance
As of November 2026, Qwen-MidTemp runs at up to 238.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
1.5B
Peak speed
238.0 tok/s
Average speed
233.3 tok/s
Avg PP
7838.5 tok/s
Min memory
3.1 GB
Max context
24,576 tokens
Avg output / run
13,440 tokens
Avg runtime / run
1m 6s
Avg quality
25.7
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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 20.1 | 25.7 | 31.3 |
| Agent Workflow | 39.8 | 56.0 | 72.2 |
| Code Generation | 0.4 | 3.8 | 7.1 |
| Role Play & Narrative | 3.1 | 13.5 | 23.9 |
| Research & Analysis | 28.8 | 29.6 | 30.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen-MidTemp 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 |
|---|---|---|---|---|---|---|---|---|
| Tesla V100-PCIE-32GB | llama.cpp | — | 238.0 tok/s | 233.3 tok/s | 3.1 GB | 24,576 tokens | 25.7 | 2 |
Benchmark runs
All 2 Qwen-MidTemp runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen-MidTemp good for agentic (tool-using) tasks?
- In our benchmarks, Qwen-MidTemp scores 56.0/100 for agentic workflows. It runs at about 233.3 tok/s, so if you want more speed, unsloth/Qwen3.6-35B-A3B-GGUF:UD-Q4_K_M is faster (~303.4 tok/s) and still scores well for agentic workflows (82.6/100).
- How fast is Qwen-MidTemp for local inference?
- Across 2 community benchmark runs, Qwen-MidTemp reaches up to 238.0 tok/s and averages 233.3 tok/s, with the fastest results on Tesla V100-PCIE-32GB.
- How much memory does Qwen-MidTemp need?
- The leanest observed configuration used about 3.1 GB of memory.
- Which tools have been used to run Qwen-MidTemp?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.