![[NEW] Supra-50M Released!](https://ai-maestro.online/wp-content/uploads/2026/05/new-supra-50m-released-1024x1024.jpg)
![[NEW] Supra-50M Released!](https://external-preview.redd.it/PFoFd4J8nvL7QW0V5dGmvPIxhnRbJ6Sa1oS-seg9nEM.png?width=140&height=75&auto=webp&s=b5e5afe0ca2df7c6510abdde1434ce7ee7b32b12 "[NEW] Supra-50M Released!")
SupraLabs released a new model! – Supra-50M
The Sproto-50M is a compact 50 million parameter causal language model (BASE and INSTRUCT versions) developed by SupraLabs using a Llama-style architecture. It was trained on 20 billion tokens of high-quality educational web text. Despite being smaller than similar open models, it achieves competitive or superior results on several key benchmarks.
What comes next?
- Sproto-124M — Base, Chat, Experimental Reasoning
- Sproto-350M — Base, Chat, Reasoning, Coding
🏆 Benchmarks
| Benchmark | Supra-50M (ours) | GPT-2 (124M) | SmolLM-135M | OpenELM-270M |
|---|---|---|---|---|
| Parameters | 50M | 124M (2.5×) | 135M (2.7×) | 270M (5.4×) |
| BLiMP (linguistics) | 76.3% | 63.0% | 69.8% | N/A |
| SciQ (science) | 77.2% | 53.2% | 73.4% | 84.70% |
| ARC-Easy (knowledge) | 52.2% | 42.0% | 49.2% | 45.08% |
| PIQA (logic) | 62.2% | 63.0% | 67.3% | 69.75% |
| HellaSwag (context) | 31.8% | 29.5% | 42.0% | 46.71% |
Source Read original →Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
