Project AURIONAI Use-Cases Research & Innovation
Stanford University
reasoning

s1

Small-data reasoning fine-tune of Qwen2.5-32B-Instruct that raised AIME24 from 26.7 to 56.7.

Stanford UniversityStanford UniversityUSAUniversity of WashingtonUniversity of WashingtonAllen AIAllen AIUSAContextual AIContextual AI
Test-time scaling chart of s1 with budget forcing, from the official GitHub README.
Levelmodel
Classreasoning
Parametersnot in the notes
Contextnot in the notes
ReleasedJan 2025
Opennessunknown
Runs onnot in the notes
Statusnot in the notes

Benchmark Results

vendor-reported
AIME24 2024math
after SFT on 1,000 examples (base 26.7)vendor-reported
56.7
050100

Score in %. Higher is better.

Jan 2025 arxiv.org
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).

Lineage

finetune
Qwen2.5-32B-Instruct to QwQ-32B-Preview: finetune, high confidenceQwen2.5-32B-Instruct to s1: finetune, high confidenceQwen2.5-32B-InstructQwen2.5-32B-InstructAlibaba (Qwen), Sep 2024QwQ-32B-PreviewQwQ-32B-PreviewAlibaba (Qwen), Nov 2024s1s1Stanford University, Jan 2025
Alibaba (Qwen)Qwen2.5-32B-Instruct is the parent (finetune)high confidencenotes
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.

R&D Topics

Small-data reasoning fine-tuning and distillationexemplar

Identity

Developer
Stanford University, University of Washington, Allen AI, Contextual AI
Family
not in the notes
Released
Jan 2025 sourcesnotes
Weights published
–
Record checked
10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
not in the notes
Total parameters
– sources
Active parameters
–
Checkpoint parameters
–
Note
Qwen2.5-32B-Instruct fine-tuned on 1,000 curated examples, 16 H100 for 26 minutes.
Context length
not in the notes
Input
not in the notes
Output
not in the notes
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
unknown
Licence
not in the notes sources
Access routes
not in the notes
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
not in the notes
Footprint
not in the notes
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
not stated
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Industry Use Cases

No industry use case links this model.

Sources

  1. s1: Simple test-time scaling arxiv.org https://arxiv.org/html/2501.19393v1

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "s1", https://aurion.semai.id/models/stanford-university-s1/, records of October 2026.