InternLM 3
InternLM3 is an open-source 8-billion parameter instruction model designed for general-purpose use and advanced reasoning. It achieves state-of-the-art performance on reasoning and knowledge-intensive tasks while being trained on only 4 trillion tokens, saving over 75% of training cost compared to similar models. It is intended for developers and researchers who need a high-performance, cost-efficient language model.
✨ Key features
- State-of-the-art reasoning and knowledge performance
- Trained on only 4 trillion tokens, saving 75%+ cost
- Supports deep thinking mode for complex reasoning
- Supports normal response mode for fluent interactions
- Available on HuggingFace, ModelScope, and Modelers
🎯 Use cases
- Solving complex reasoning tasks with long chain-of-thought
- Building conversational AI applications
- Knowledge-intensive question answering
- General-purpose language understanding and generation
⚠️ Good to know
The model may produce unexpected outputs including biases or harmful content due to its probabilistic nature; users should not propagate such content.
❓ FAQ
What is the model size of InternLM3?
InternLM3-8B-Instruct is an 8-billion parameter instruction model.
How does InternLM3 achieve cost savings?
It is trained on only 4 trillion high-quality tokens, saving more than 75% of training cost compared to other LLMs of similar scale.
What modes does InternLM3 support?
It supports deep thinking mode for complex reasoning via long chain-of-thought and normal response mode for fluent interactions.
Where can I download the model?
The model is available on HuggingFace, ModelScope, and Modelers.
📊 Repository
🤖 Overview, features, install steps and FAQ were generated from the project's README on Sep 4, 2026. Always check the original source before running commands.