Model benchmarks
iq3_xxs local LLM performance
As of November 2026, iq3_xxs runs at up to 119.9 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
119.9 tok/s
Average speed
117.4 tok/s
Avg PP
599.3 tok/s
Min memory
n/a
Max context
131,072 tokens
Avg output / run
74,465 tokens
Avg runtime / run
10m 34s
Avg quality
82.2
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 | 80.5 | 82.2 | 84.0 |
| Agent Workflow | 86.6 | 88.0 | 89.3 |
| Code Generation | 47.2 | 56.0 | 64.8 |
| Role Play & Narrative | 95.0 | 95.1 | 95.2 |
| Research & Analysis | 89.4 | 89.8 | 90.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination iq3_xxs 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3090 | Strata | — | 119.9 tok/s | 117.4 tok/s | n/a | 131,072 tokens | 82.2 | 2 |
Benchmark runs
All 2 iq3_xxs runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is iq3_xxs good for coding?
- In our benchmarks, iq3_xxs scores 56.0/100 for coding. It runs at about 117.4 tok/s, so if you want more speed, Qwen3.6-35B-A3B-Q4_K_XL is faster (~156.9 tok/s) and still scores well for coding (82.2/100).
- Is iq3_xxs good for agentic (tool-using) tasks?
- In our benchmarks, iq3_xxs scores 88.0/100 for agentic workflows. It runs at about 117.4 tok/s, so if you want more speed, swift-1.5-qwen3.8-flash-next-iq3_s is faster (~259.0 tok/s) and still scores well for agentic workflows (88.9/100).
- How fast is iq3_xxs for local inference?
- Across 2 community benchmark runs, iq3_xxs reaches up to 119.9 tok/s and averages 117.4 tok/s, with the fastest results on NVIDIA GeForce RTX 3090.
- Which tools have been used to run iq3_xxs?
- Benchmarks were submitted using Strata. Results are community-contributed and updated as new runs arrive.