Modern tools for AI research
As ASU researchers set out to solve humanity’s greatest challenges, our goal is to provide the best tools to empower them to succeed. AI is an essential part of the modern toolkit, offering new opportunities to advance research that matters to our communities today.
Knowledge Enterprise RTO and Research Computing at ASU provide world-class AI models, tools and high-performance computing resources to faculty, students, staff and community members.
By opening access to tools and platforms across all disciplines, we foster an inclusive environment for AI research that joins scientific, humanistic and artistic exploration.
Learn more about ASU’s AI Research Acceleration (AIR) Platform and request a consultation.
Our infrastructure, which includes the Sol supercomputer, offers the power and flexibility to push boundaries in AI research — whether you are developing large language models (LLMs), accelerating AI-driven software, or applying deep learning techniques to complex problems.
Boosting research with AI tools
Studying and sorting through large datasets is an important part of research. This is exactly where AI models offer a major advantage — they are ideally suited to analyze huge amounts of information with speed.
Research Computing provides local access to powerful open-source AI models on the Sol supercomputer, displayed in the table below. Researchers can use these models separately or in combination to examine datasets stored locally. This improves data security by eliminating the need for cloud uploads.
Sol also carries a suite of AI-powered tools, including MONAI, AlphaFold3, TensorFlow and PyTorch. These provide faculty, students and staff with a robust computing environment for AI-enabled research.
ASU’s advanced computing resources not only drive innovation and discovery in AI research, they contribute to numerous scientific publications across various disciplines. Researchers leveraging ASU’s Sol supercomputer, local AI models and GPU acceleration have produced impactful studies in areas such as machine learning, large language models, biomedical AI and computational science.
Powering research
that matters
ASU faculty, staff and student researchers are leveraging the power of AI for public good. With access to cutting-edge technologies, they are accelerating research and broadening impact in areas such as health care, energy, water, national security and more. They are also studying AI itself — applying a wide range of disciplines to improve the technology’s performance, security, reliability, accessibility and sustainability.
Choosing the right AI model: What to know
Before diving into the model comparisons, here are terms to help interpret the table:
- Parameters: Parameters are the “brainpower” of an AI model. The more parameters a model has, the more information it can process and the more complex patterns it can learn.
- Context window: This refers to how much text a model can consider at once. A larger context window allows the model to understand and respond to longer inputs. This is useful for analyzing documents, conducting in-depth conversations or performing multi-step reasoning.
Open-source AI models on the Sol computer
| Model | Release date | Specialties and features | Parameters | Context window |
| Whisper | September 2022 | Multilingual speech-to-text models optimized for fast and accurate open-source transcription | 39M (Tiny), 244M (Small), 769M (Medium), 1.55B (Large) | 30 sec audio |
| Qwen3-Coder-30B-A3B-Instruct | March 2026 | Frontier coding and agentic software-engineering model optimized for SWE-bench and repository reasoning | 30B MoE (~3B active) | 256k |
| OLMo 3.1 32B Instruct | February 2026 | Fully open research-oriented reasoning and instruction-following model trained on transparent datasets | 32B | 128k |
| Gemma 4 | March 2026 | Frontier multimodal instruction-tuned family optimized for reasoning, coding and efficient local deployment | 26B A4B MoE (~4B active), 31B | 256k |
| DeepSeek V4 Flash | January 2026 | Frontier reasoning and coding model with optimized inference and agentic workflows | MoE class | 128k–256k |
| Qwen3.6-27B | April 2026 | State-of-the-art general-purpose open model for coding, reasoning, multilingual tasks and tool use | 27B | 128k |
| Qwen3-VL-2B Instruct | March 2026 | Lightweight vision-language model for OCR, document understanding and multimodal assistants | 2B | 128k |
| MiniMax-M2 | April 2026 | Frontier agentic and coding-focused mixture-of-experts model with strong tool-use performance | MoE class | 200k |
Model selection tips
The best model for your research depends on your needs. Here are some general guidelines:
- For speech-to-text and transcription, use the Whisper models. Smaller versions are faster, while Medium and Large provide higher accuracy.
- For coding and software engineering, Qwen3-Coder-30B-A3B-Instruct is optimized for code generation, reasoning and agentic workflows.
- For general-purpose reasoning and research assistance, Qwen3.6-27B offers a strong balance of performance, multilingual support and efficient deployment.
- For multimodal tasks such as OCR, document understanding and image analysis, use Qwen3-VL-2B Instruct or the Gemma 4 family.
- For advanced reasoning and long-context workflows, DeepSeek V4 Flash and MiniMax-M2 provide frontier-scale capabilities but require more GPU resources.
- For fully open and reproducible AI research, OLMo 3.1 32B Instruct emphasizes transparency in datasets and training methodology.
- If you are unsure where to start, Qwen3.6-27B is a strong default choice for many research applications.
A safe AI environment
Many scientists want to leverage AI but also need to protect proprietary data. The university’s OpenAI API, offered through Research Computing, empowers ASU researchers to integrate OpenAI’s advanced models into their projects while ensuring their data remains private and owned by ASU.
Getting started
All AI models listed on this page are freely available for ASU researchers on the Sol supercomputer. To use the AI models, you must first request access to the Sol supercomputer. Faculty and research staff may request accounts directly. Students and non-faculty users must be sponsored by an ASU faculty member. Once your account is approved, you will receive onboarding instructions and access credentials.
Next, to begin using the models, schedule a consultation with the Research Computing team. We’ll walk you through accessing the Sol supercomputer via the web portal, launching a Jupyter Notebook and selecting the appropriate tools for your research. This personalized guidance will help ensure you’re set up for success — especially if you’re new to working with AI models in a supercomputing environment.
Events
RTO provides cutting-edge resources, training and events to empower researchers in the rapidly evolving field of AI.
Bringing AI to the classroom
ASU equips students to engage in hands-on AI learning, giving them valuable experience with technology that continues to shape the modern workforce. Students can access AI applications on the Sol supercomputer if their instructors request student accounts from Research Computing, which provides AI computing support for academic courses.
Students with accounts can immediately begin working with cutting-edge AI tools without instructors needing to handle software setup. Research Computing pre-configures software environments, including several Python environments and ready-to-use Jupyter notebooks integrated with AI models.
Below is a sample of past AI-related courses supported by Research Computing.
| Course number | Course name |
| EEE 549 | Statistical Machine Learning: From Theory to Practice |
| CSE 575 | Statistical Machine Learning |
| CSE 598 | Frontier topics in GenAI |
| EEE 598 | Deep Learning: Foundations and Applications |
| CSE 576 | Topics in Natural Language Processing |
| EEE 598 | Generative AI: Theory and Practice |
| CIS 508 | Machine Learning in Business |
| CEN 524, CSE 524, CSE 494 | Machine Learning Acceleration |
| CSE 476 | Introduction to Natural Language Processing |
| MFG 598 | AI in Additive Manufacturing |
| FIN 597 | AI and Machine Learning Applications in Finance |
Not sure what you need?
Fill out our general request help form for a consultation with a research facilitator.