---
license: cc-by-4.0
language:
- en
size_categories:
- n<1K
task_categories:
- text-generation
tags:
- benchmark
- inferencebench
- llm
- microsoft
- h100
inferencebench:
  envelope_version: v1
  suite_id: llm.quality.arithmetic-mini
  suite_version: 1.0.0
  model: microsoft/Phi-3.5-mini-instruct
  engine: vllm
  hardware_class: h100
  fingerprint_sha256: 550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7
  signature_verified: false
  rekor_log_index: -1
---

# microsoft/Phi-3.5-mini-instruct on llm.quality.arithmetic-mini (NVIDIA H100 80GB HBM3)

[Back to leaderboard](https://huggingface.co/spaces/yobitel/leaderboard-llm.quality.arithmetic-mini)

## Headline metrics

| Metric | Value | Unit |
|---|---|---|
| N Samples | 8 |  |
| N Ok | 8 |  |
| Ok Rate | 1 |  |
| Accuracy | 0.75 |  |
| Accuracy P05 | 0 |  |
| Accuracy P50 | 1 |  |
| Accuracy P95 | 1 |  |
| TTFT P50 | 11.2417 | ms |
| Total P50 Ms | 31.7159 |  |
| Tokens Out Total | 208 |  |

## Run configuration

- **Model**: microsoft/Phi-3.5-mini-instruct @ unknown00
- **Engine**: vllm vunknown
- **Quantization**: fp16
- **Hardware**: NVIDIA H100 80GB HBM3
- **Driver**: 580.126.09
- **CUDA**: 13.0
- **Run date**: 2026-05-18T13:53:40.482136+00:00
- **Seed**: 0

## Verification

This result is Sigstore-signed and Rekor-logged. Verify:

```bash
pip install inferencebench
bench verify hf://datasets/Yobitel/microsoft-phi-3-5-mini-instruct__llm-quality-arithmetic-mini__019e3b5d2482/envelope.json
```

[Rekor entry: log index -1](https://search.sigstore.dev/?logIndex=-1)

## Methodology

See [the suite methodology page](https://yobitelcomm.github.io/bench/suites/llm.quality.arithmetic-mini).

## Citation

```bibtex
@misc{inferencebench_019e3b5d2482,
  title = { microsoft/Phi-3.5-mini-instruct on llm.quality.arithmetic-mini },
  author = { {InferenceBench community} },
  year = { 2026 },
  url = { https://huggingface.co/datasets/Yobitel/microsoft-phi-3-5-mini-instruct__llm-quality-arithmetic-mini__019e3b5d2482 },
}
```

---

*Published via [InferenceBench](https://github.com/yobitelcomm/bench) — vendor-neutral AI benchmarks.*