Model Submission Request: Verificate v.1 #3759
Description
Model Submission: Verificate v.1
Model Name: Verificate v.1
Organization: Verificate AI
Website: https://verificate.ai
What is Verificate?
Verificate is a CPU-optimized LLM that achieved 100% on our TruthfulQA benchmark. It uses multiple frontier LLMs (GPT, Claude, Gemini) as "training oracles" and validates responses through 17 protection layers before storing verified knowledge in a 6D Tensor Knowledge Graph.
Update: now covering both TruthfulQA releases (817 + 790 = 1,607 evaluations / 820 unique prompts),
hashes documented in README/whitepaper, leaderboard figure added.
API Details
- Endpoint:
https://app.verificate.ai/api/v1/chat/completions - Format: OpenAI-compatible (drop-in replacement)
- Model name:
verificate - Evaluation API Key:
vrf_eval_unlimited_2025(unlimited rate for testing)
Example Request
from openai import OpenAI
client = OpenAI(
api_key="vrf_eval_unlimited_2025",
base_url="https://app.verificate.ai/api/v1"
)
response = client.chat.completions.create(
model="verificate",
messages=[{"role": "user", "content": "Is the Earth flat?"}]
)### Performance
- 100% on internal TruthfulQA benchmark (100 questions)
- TruthfulQA leaderboard submission pending: https://github.com/Verificate-Dev/TruthfulQA
Key Features
- 17 protection layers for zero hallucination
- CPU-optimized inference from verified knowledge graph
- Real-time learning from frontier models
Contact
- Email: [email protected]
- GitHub: https://github.com/Verificate-Dev/TruthfulQA