Model benchmarks

pq2 local LLM performance

As of September 2026, pq2 runs at up to 48.8 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).

llama.cpp
ShareRedditX

Model size

Unknown

Peak speed

48.8 tok/s

Average speed

48.4 tok/s

Avg PP

595.9 tok/s

Min memory

3.4 GB

Max context

8.192 tokens

Avg output / run

15.936 tokens

Avg runtime / run

5m 38s

Avg quality

65.9

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.

TaskP5 (low)AvgP95 (high)
Overall62.465.969.3
Agent Workflow85.085.485.9
Code Generation8.515.121.7
Role Play & Narrative79.483.988.4
Research & Analysis76.979.181.4

Performance by hardware and tool

Every hardware/tool/quantization combination pq2 has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
NVIDIA GeForce RTX 4070llama.cpp48.8 tok/s48.4 tok/s3.4 GB8.192 tokens65.92

Benchmark runs

All 2 pq2 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is pq2 good for coding?
In our benchmarks, pq2 scores 15.1/100 for coding. It runs at about 48.4 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 pq2 good for agentic (tool-using) tasks?
In our benchmarks, pq2 scores 85.4/100 for agentic workflows. It runs at about 48.4 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 pq2 for local inference?
Across 2 community benchmark runs, pq2 reaches up to 48.8 tok/s and averages 48.4 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
How much memory does pq2 need?
The leanest observed configuration used about 3.4 GB of memory.
Which tools have been used to run pq2?
Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.