Happy Horse 1.0 is now on ModelsLab

Try Now
Skip to main content
Available now on ModelsLab · Language Model

Prime Intellect: INTELLECT-3Reason Deeper. Scale Smarter

Try Prime Intellect: INTELLECT-3API Documentation

Master Math. Code. Agents

106B MoE

Efficient Active Params

12B active parameters deliver math and reasoning at low inference cost.

RL Post-Training

End-to-End Reinforcement

Trained with prime-rl on SFT and RL from GLM-4.5-Air for superior agentic performance.

Benchmark Dominance

Outperforms Larger Models

Hits 90.8% AIME 2024, beats 6x bigger models on reasoning and code tasks.

Examples

See what Prime Intellect: INTELLECT-3 can create

Copy any prompt below and try it yourself in the playground.

Math Proof

“Prove the fundamental theorem of calculus step-by-step, including all intermediate derivations and edge cases for rigorous validation.”

Code Optimizer

“Write efficient Python code to solve the traveling salesman problem using dynamic programming with memoization, include time complexity analysis.”

Science Hypothesis

“Formulate a testable hypothesis for quantum entanglement effects in macroscopic systems, design experiment with controls and measurements.”

Agent Workflow

“Design multi-step agent plan to analyze market data, fetch real-time stocks via API, compute trends, and generate trading signals with risk assessment.”

For Developers

A few lines of code.
RL Agents. One Call

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

  • Serverless: scales to zero, scales to millions
  • Pay per token, no minimums
  • Python and JavaScript SDKs, plus REST API
API Documentation 
import requests


response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
  "key": "YOUR_API_KEY",
  "prompt": "",
  "model_id": ""
}
)
print(response.json())

FAQ

Common questions about Prime Intellect: INTELLECT-3

Read the docs

Prime Intellect: INTELLECT-3 is a 106B-parameter Mixture-of-Experts LLM with 12B active params. It excels in math, code, science, and reasoning benchmarks. Built on GLM-4.5-Air with SFT and RL.

prime intellect intellect 3 API provides LLM endpoint for inference. Supports 131k context for complex tasks. Use prime-rl stack for custom training.

Achieves 90.8% on AIME 2024, 88.0% AIME 2025, 69.3% LiveCodeBench. Outperforms larger models like DeepSeek and matches GLM-4.6.

Yes, full model weights, prime-rl framework, datasets, and environments are open-sourced. Access via Prime Intellect platform.

Serves as open alternative to closed models like GPT-4 or Claude. Beats many frontier LLMs over 6x larger on agentic tasks.

Supports 131k-token context window. Enables long-form reasoning and multi-turn agent interactions without truncation.

Ready to create?

Start generating with Prime Intellect: INTELLECT-3 on ModelsLab.

Try Prime Intellect: INTELLECT-3API Documentation