P
Prime Intellect: INTELLECT-3
● Excellent8.3
/ 10
INTELLECT-3 is a 106B-parameter Mixture-of-Experts model (12B active) post-trained from GLM-4.5-Air-Base using supervised fine-tuning (SFT) followed by large-scale reinforcement learning (RL). It offers state-of-the-art performance for its size across math,...
Specifications
| Attribute | Value |
|---|---|
| Lab | Prime Intellect |
| Tags | Open Weight Intelligent |
| Overall Score | 8.3/10 |
| Release Date | 2025-11 |
| Context Window | 131,072 tokens |
| Input Price / 1M | $0.20 |
| Output Price / 1M | $1.10 |
| Input Modalities | Text |
| Output Modalities | Text |
Strengths
- 106B total parameters with 12B active (MoE architecture)
- Post-trained from GLM-4.5-Air-Base with SFT + large-scale RL
- State-of-the-art performance for its size class
- Budget pricing at $0.20/$1.10 per 1M tokens
- Strong across reasoning, coding, and general intelligence benchmarks
Weaknesses
- Derived from GLM-4.5 base model, not trained from scratch
- Smaller community and ecosystem vs Qwen/DeepSeek alternatives
- Limited deployment options outside of Prime Intellect/OR
Best For
- Cost-effective high-quality inference
- Open-weight fine-tuning and customization
- General reasoning and coding tasks
- Research and experimentation with MoE architectures
Sources & Further Reading
Related Models
Scores are aggregated from public benchmarks (MMLU, HumanEval, MATH, GSM8K, LMSYS) and normalized to a 1–10 scale. Methodology →