# Soofi S 30B-A3B Benchmark Scores & Performance

> Soofi S 30B-A3B by Soofi Project has 8 source-displayable benchmark rows but no public overall score. It remains unranked.

Canonical page: https://benchlm.ai/models/soofi-s-30b-a3b

Last updated: October 14, 2026

## Model Details

| Property | Value |
|----------|-------|
| Creator | Soofi Project |
| Source Type | Open Weight |
| Reasoning Type | Non-Reasoning |
| Context Window | 1M |
| Overall Score | Not computed (unranked) |
| Overall Rank | Unranked |

## Family & Coverage

- Family: Soofi S
- Variant: base (30B-A3B)
- Benchmarks covered: 8 of 628
- Coverage note: Published evidence is shown below; no public overall score is assigned.

## Coding Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [HumanEval](/benchmarks/humaneval) | 73.8% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |

## Reasoning Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [BBH](/benchmarks/bbh) | 78.8% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |
| [DROP](/benchmarks/drop) | 66.5% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |

## Knowledge Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [GPQA](/benchmarks/gpqa) | 43.4% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |
| [GPQA-D](/benchmarks/gpqa-diamond) | 43.4% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |
| [MMLU-Pro](/benchmarks/mmlu-pro) | 51.4% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |
| [AGIEval](/benchmarks/agieval) | 66.9% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |

## Mathematics Benchmarks

| Benchmark | Score | Source |
|-----------|-------|--------|
| [GSM8K](/benchmarks/gsm8k) | 86.1% | [Soofi Project: Soofi S pretraining report](https://arxiv.org/abs/2607.09424) |