Skip to content
AI Model Radar

internlm3 8b instruct

internlm/internlm3-8b-instruct
Open Source · chat · open-weights
Open Alert me on changes
Context
—
Max output
—
Weights
Open
API $/1M
—
Modalities
text
Released
13 Jan 2025
License: apache-2.0 · internlm/internlm3-8b-instruct
Download image Share on X Share on LinkedIn
AI summary
● machine-written

InternLM3-8B-Instruct: 8B open-source model for reasoning and chat

InternLM3-8B-Instruct is an open-source 8-billion parameter instruction-tuned model designed for general-purpose usage and advanced reasoning tasks. The model supports both standard conversation mode and a deep thinking mode for complex reasoning, and demonstrates competitive performance on benchmarks including MMLU-Pro (57.6), reasoning tasks, and code generation. Training used only 4 trillion high-quality tokens, reducing training costs by over 75% compared to similar-scale models.

What's new
  • Supports deep thinking mode for long chain-of-thought reasoning on complex tasks
  • Achieves 57.6 on MMLU-Pro, 83.1 on MATH-500, and 82.3 on HumanEval (Pass@1)
  • Trained on 4 trillion tokens with claimed 75% cost reduction vs. similar models
  • Long context performance: 87.9 average on RULER (4-128K token evaluation)
Best for
Complex reasoning and problem-solving with thinking modeGeneral-purpose chat and instruction followingKnowledge-intensive and mathematical tasksCode generation and understanding
Sources

Source: https://huggingface.co/internlm/internlm3-8b-instruct