Guides & Tutorials: Harnessing GenAI in Your Academic Journey: Chatbot Tools

Chatbots utilizing different large language models (LLMs) for various language tasks, such as refining research questions, checking grammar, polishing language, and translation, or even enhanced information search.

PolyU GenAI App

PolyU ITS offers a variety of GenAI chatbots for staff and students to use in teaching, research, and work-related activities. Access requires authentication with a PolyU NetID. For technical issues of this service, please contact ITS.

The following compares these available GenAI large language models (LLMs), which can be used for multiple purposes in research.

Please note that the information in this section is subject to revision and may change over time.
Last updated: Apr, 2025

PolyU students and staff have 1850 credits per month to use across various LLMs and image generation tools.

The first table compares the 4o series, O1 series, and DeepSeek, which all consume credits.

For quick and generic tasks -- Use the 4o series or the credit-free models, compared in the second table.
For complex reasoning -- Use O1 series or DeepSeek perform better, though they consume more credits and take longer to process.

Model	Release	Total Parameters	Key Features	Max. Prompt Size (in English Characters)	Host	Credit Consumption
GPT-4o (OpenAI)	May 2024	Not disclosed	Fast processing Supports image-to-text Supports more complex text prompts Better translation between languages	100,000	Microsoft Azure Cloud	*GPT-4o consumes more credit than GPT-4o-mini
GPT-4o-mini (OpenAI)	Jul 2024		Supports image-to-text Scores lower on some benchmarks compared to GPT-4o	100,000		*GPT-4o consumes more credit than GPT-4o-mini
O1 (OpenAI) NEW	Dec 2024		Capable of breaking down tasks logically without detailed guidance by using an internal chain of thought Takes longer time to reason and respond Excels in complex reasoning tasks, esp. in science, coding, and math O3-mini is weaker in factual knowledge on non-STEM topics, only comparable to smaller LLMs like 4o-mini O1 can reason over images, but not O3-mini	700,000		Input: 6x GPT-4o Output: 6x GPT-4o Internal chain of thought also consumes tokens
O3-mini (OpenAI) NEW	Jan 2025			700,000		Input: 20x GPT-4o-mini Output: 20x GPT-4o-mini Internal chain of thought also consumes tokens
Deepseek-R1 (Azure AI Foundry) (DeepSeek AI) NEW	Jan 2025	671B, with 37B active parameters	Performance comparable to O1 models across math, coding, and reasoning tasks Capable of analyzing your problem using an observable chain of thought before responding	400,000	Azure AI Foundry	Input: 1.35 credits/1,000 tokens Output: 5.40 credits/1,000 tokens Observable chain of thought also consumes tokens
DeepSeek-R1 (Alibaba Cloud) (DeepSeek AI) NEW	Jan 2025	671B, with 37B active parameters		200,000	Alibaba Cloud	Input: 0.55 credits/1,000 tokens Output: 2.20 credits/1,000 tokens Observable chain of thought also consumes tokens

The use of the models below will NOT consume the monthly use entitlement credit, however, please be aware that for Copilot with Bing, there are 30 response quotas for each conversation and a new conversation can be initiated once 30 responses have been reached.

Model	Release	Total Parameters	Key Features	Max. Prompt Size (in English Characters)	Host
Qwen2.5-72B-Instruct 通义千问 (Alibaba)	Sep 2024	72B	Excels in conversational understanding and math problems Enhanced multilingual support, especially for tasks in Chinese	100,000	PolyU Campus^
Qwen2.5-VL-72B-Instruct 通义千问 (Alibaba) NEW	Jan 2025	72B	Excels in visual understanding, e.g. recognize charts, icons, graphics and layouts within images Excel in text analysis
Llama 3.3-70B Instruct (Meta)	Dec 2024	70B	Optimized for multilingual dialogue use cases Supports English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai
Mistral-Large-Instruct-2407 (Mixtral AI)	Jul 2024	123B	Excellent for complex reasoning Comparable to GPT-4o in common sense and reasoning Good at processing code inputs
Copilot with Bing (Microsoft)	Nov 2023?	Not disclosed	Integrated with Bing search for real-time information retrieval Supports image-to-text Can be installed in Edge, computers, and mobile devices Powered by GPT-4, Prometheus, and DALL-E3?	16,000	Service provided by Microsoft (Copilot Commercial Data Protection (CDP) available c.f. personal account)

^Information input to GenAI models hosted in PolyU Campus is encrypted and will be erased within 48 hours.

	2766-6863
	2766-6863 (service hours)
	Online Form
	Contact your Faculty Librarians on in-depth research questions