Model benchmarks
swift-iq2_xs local LLM performance
As of November 2026, swift-iq2_xs runs at up to 92.9 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
92.9 tok/s
Average speed
91.5 tok/s
Avg PP
524.8 tok/s
Min memory
259 MB
Max context
32,768 tokens
Avg output / run
56,235 tokens
Avg runtime / run
10m 1s
Avg quality
87.9
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 | 87.4 | 87.9 | 88.5 |
| Agent Workflow | 91.3 | 91.4 | 91.4 |
| Code Generation | 77.1 | 78.7 | 80.2 |
| Role Play & Narrative | 89.6 | 92.8 | 96.1 |
| Research & Analysis | 88.4 | 88.8 | 89.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination swift-iq2_xs 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 4070 | Strata | — | 92.9 tok/s | 91.5 tok/s | 259 MB | 32,768 tokens | 87.9 | 2 |
Benchmark runs
All 2 swift-iq2_xs runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is swift-iq2_xs good for coding?
- In our benchmarks, swift-iq2_xs scores 78.7/100 for coding. It runs at about 91.5 tok/s, so if you want more speed, Ornith-1.5-35B-A3B.APEX-I-NanoPlus is faster (~110.5 tok/s) and still scores well for coding (81.1/100).
- Is swift-iq2_xs good for agentic (tool-using) tasks?
- In our benchmarks, swift-iq2_xs scores 91.4/100 for agentic workflows, among the top 2 for agentic workflows on consumer hardware (≤24 GB VRAM). It runs at about 91.5 tok/s, so if you want more speed, ornith-1.5:35B is faster (~115.1 tok/s) and still scores well for agentic workflows (88.4/100).
- How fast is swift-iq2_xs for local inference?
- Across 2 community benchmark runs, swift-iq2_xs reaches up to 92.9 tok/s and averages 91.5 tok/s, with the fastest results on NVIDIA GeForce RTX 4070.
- How much memory does swift-iq2_xs need?
- The leanest observed configuration used about 259 MB of memory.
- Which tools have been used to run swift-iq2_xs?
- Benchmarks were submitted using Strata. Results are community-contributed and updated as new runs arrive.