Exploring AI Models

The AI Models section provides access to a wide range of pre-trained and custom models. These models combine instructions, additional knowledge, and specialized skills for various tasks.

To get started, navigate to the Models tab.

What you can do on the Models Dashboard:

• Browse the list of models with their context lenghts and prices.

Popular AI Models

Model Name	Developer	Description
deepseek-ai/DeepSeek-R1-0528	Deepseek	Enhanced model with improved reasoning, inference, and algorithmic post-training optimizations; designed for high-accuracy tasks.
meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8	Meta AI	Multimodal instruction-tuned model leveraging a mixture-of-experts (MoE) architecture for top-tier performance in both text and image understanding.
gpt-oss-120b	Open AI	Open-weight 117B parameter Mixture-of-Experts model supporting 128k context, advanced reasoning via chain-of-thought, optimized for real-world tool use, coding, and efficient local or cloud deployment.
Intel/Qwen3-Coder-480B-A35B-Instruct-int4-mixed-ar	Qwen	High-capacity, instruction-tuned code generation model optimized with INT4 mixed-precision for fast inference, designed for complex programming tasks on Intel hardware.
Qwen3-Next-80B-A3B-Instruct	Qwen	High-capacity model optimized for instruction following and knowledge-intensive tasks.
gpt-oss-20b	Open AI	Open-source GPT-style model suitable for text generation and general-purpose tasks.
Qwen3-235B-A22B-Thinking-2507	Qwen	Powerful 235B-parameter language model optimized for deep reasoning, planning, and complex multi-step tasks.
Mistral-Nemo-Instruct-2407	Mistral AI	Instruction-tuned model focusing on efficient reasoning and NLP tasks.
meta-llama/Llama-3.3-70B-Instruct	Meta AI	Large-scale transformer model fine-tuned for instruction-following, aligning responses with human preferences.
mistralai/Mistral-Large-Instruct-2411	Mistral AI	Large instruction-tuned model offering strong general-purpose reasoning, summarization, and assistant-style responses.
Qwen/Qwen2.5-VL-32B-Instruct	Qwen	Powerful vision-language model trained to follow multimodal instructions, suitable for image understanding, captioning, and reasoning.
meta-llama/Llama-3.2-90B-Vision-Instruct	Meta AI	Vision-language model with instruction tuning, capable of image analysis, visual Q&A, and multimodal dialogue generation.
BAAI/bge-multilingual-gemma2	BAAI	Multilingual embedding model optimized for semantic search and retrieval tasks across diverse languages.
zai-org/GLM-4.6	Z.AI	Advanced large-language model that expands context capacity to 200K tokens and significantly enhances coding, reasoning, and agentic capabilities. It excels in real-world coding tools, delivering more natural, human-aligned outputs.
zai-org/GLM-4.7	Z.AI	Advanced large-language model that retains a 200K-token context window and elevates coding, reasoning, and agentic capabilities with enhanced multi-step execution and consistency. It introduces sophisticated thinking modes like Preserved Thinking and Turn-level Thinking.
moonshotai/Kimi-K2-Instruct-0905	Moonshot AI	A state-of-the-art mixture-of-experts (MoE) language model, featuring 32 billion activated parameters and a total of 1 trillion parameters. It delivers exceptional reasoning, coding, and content-generation performance.
moonshotai/Kimi-K2-Thinking	Moonshot AI	A high-performance open-source thinking model built for step-by-step thinking and dynamic tool use. It achieves state-of-the-art results on benchmarks such as Humanity’s Last Exam (HLE) and BrowseComp by dramatically scaling multi-step reasoning depth maintaining stable tool-use across 200–300 sequential calls.
deepseek-ai/DeepSeek-V3.2	Deepseek	A LLM model that combines breakthrough efficiency with exceptional reasoning and tool-use performance. Powered by Sparse Attention and scalable RL post-training, it delivers premium long-context quality at reduced cost. Reports place it in the GPT-5 class with gold-medal wins in the 2025 IMO and IOI.

For a full list of models, visit the AI Models section in your dashboard.

Testing an AI Model

Before using any of our AI models in your project, you can perform real-time testing directly from the dashboard. This allows you to evaluate the model’s performance and ensure it meets your requirements.

How to Test an AI Model

1. Select a Model:
   • Navigate to the Models tab.
   • Select the model you want to test by clicking on it.
2. Start Testing:

   • On the AI model chat page, type your question or input into the centered text field.

   • Press your Enter key or the arrow icon to submit your query.

     
     Your Daily Credits usage is shown above the request field. To view detailed model-specific usage information, visit the IO Intelligence Payments page.

3. Interact with the Model:
   • The AI model will respond, starting a conversation. You can continue testing by asking additional questions or providing more input.
   • Compare different models to find the one that best suits your needs.

Managing Chats

• View Previous Chats:
  • On the left, you can manage your previously created chats with different AI models.
  • Click on any chat to dive deeper into the conversation.
• Create or Remove Chats:

  • Create a new chat by clicking the + button next to the model name.

  • Remove a chat by clicking the three-dot menu and selecting Delete. Remove unnecessary chats to keep your workspace organized.

    

Switching Between Models

• Next to the + button is a dropdown menu showing the currently selected model.

• Click the dropdown menu to select a different AI model and begin a new conversation with it.