Model benchmarks
iq3_s local LLM performance
As of October 2026, iq3_s runs at up to 56.1 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
Unknown
Peak speed
56.1 tok/s
Average speed
48.9 tok/s
Avg PP
332.4 tok/s
Min memory
n/a
Max context
65,536 tokens
Avg output / run
72,066 tokens
Avg runtime / run
24m 49s
Avg quality
83.5
Benchmark runs
2
GPUs tested
2
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 | 82.1 | 83.5 | 84.9 |
| Agent Workflow | 84.4 | 87.5 | 90.7 |
| Code Generation | 61.9 | 64.7 | 67.5 |
| Role Play & Narrative | 92.5 | 93.9 | 95.2 |
| Research & Analysis | 86.4 | 88.1 | 89.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination iq3_s has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 7900 XTX | OpenAI-compatible | — | 56.1 tok/s | 56.1 tok/s | n/a | 65,536 tokens | 85.1 | 1 |
| NVIDIA GeForce RTX 3060 | OpenAI-compatible | — | 41.6 tok/s | 41.6 tok/s | n/a | 65,536 tokens | 82.0 | 1 |
Benchmark runs
All 2 iq3_s runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is iq3_s good for coding?
- In our benchmarks, iq3_s scores 64.7/100 for coding. It runs at about 48.9 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is iq3_s good for agentic (tool-using) tasks?
- In our benchmarks, iq3_s scores 87.5/100 for agentic workflows. It runs at about 48.9 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 iq3_s for local inference?
- Across 2 community benchmark runs, iq3_s reaches up to 56.1 tok/s and averages 48.9 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
- Which tools have been used to run iq3_s?
- Benchmarks were submitted using OpenAI-compatible. Results are community-contributed and updated as new runs arrive.