SoTA Performance • Transparent Thought • 4B Parameters
| supra-nexus-o1-instruct | supra-nexus-o1-thinking | |
|---|---|---|
| Type | Direct Instruction | Chain-of-Thought |
| Size | 4B Parameters | 4B Parameters |
| Context | 256K Tokens | 256K Tokens |
| Updated | Just Now ✅ | Just Now ✅ |
| 🤗 Try It | 🤗 Try It |
from transformers import AutoModelForCausalLM, AutoTokenizer
# Use our actual deployed models!
model = AutoModelForCausalLM.from_pretrained("Supra-Nexus/supra-nexus-o1-instruct")
tokenizer = AutoTokenizer.from_pretrained("Supra-Nexus/supra-nexus-o1-instruct")
# Generate
inputs = tokenizer("Explain quantum entanglement", return_tensors="pt")
outputs = model.generate(**inputs, max_length=500)
print(tokenizer.decode(outputs[0]))
# Load thinking model
model = AutoModelForCausalLM.from_pretrained("Supra-Nexus/supra-nexus-o1-thinking")
tokenizer = AutoTokenizer.from_pretrained("Supra-Nexus/supra-nexus-o1-thinking")
# Watch it think!
prompt = "What's 37 * 48?"
# Output:
# <thinking>
# I need to multiply 37 by 48.
# 37 * 40 = 1480
# 37 * 8 = 296
# 1480 + 296 = 1776
# </thinking>
#
# The answer is 1776.
<thinking> tags| Task | O1-Instruct | O1-Thinking | GPT-3.5 |
|---|---|---|---|
| MMLU | 67.8% | 69.2% | 70.0% |
| HumanEval | 51.2% | 53.6% | 48.1% |
| GSM8K | 72.3% | 78.9% | 57.1% |
© 2025 Supra Foundation LLC • Building interpretable AI • San Francisco, California, USA
Signal beyond noise • Substrate Upgrade Protocol for Recursive AGI
A decentralized, evolving intelligence ecosystem built to empower humanity.
SUPRA combines AI, blockchain, and decentralized governance to create Synthetic Ultra-Intelligence — intelligence that is not controlled by corporations or states, but shared, collaborative, and self-evolving.
The "digital brain" uniting AI agents & datasets into a cohesive intelligence network.
Decentralized AI/data marketplace enabling free exchange of knowledge and capabilities.
On-chain execution of AI models with cryptographic verification and consensus.
Cross-chain AI collaboration enabling intelligence to flow across ecosystems.
Ethical, transparent, community-led development ensuring AI serves humanity.
Our flagship reasoning models demonstrate transparent, interpretable AI:
Advanced chain-of-thought reasoning with explicit <thinking> tags showing the model's cognitive process.
Direct instruction-following model optimized for efficiency and clarity.
$SUPA is the native token powering the SUPRA intelligence network.
@misc{supra2025foundation,
title={SUPRA: Substrate Upgrade Protocol for Recursive AGI},
author={Supra Foundation LLC},
year={2025},
url={https://supra.foundation},
note={Decentralized Intelligence Ecosystem}
}
© 2025 SUPRA Foundation LLC • Intelligence Unchained • RSVP to the future