Spaces:
Running on CPU Upgrade
Cost + latency + hallucination evaluation for Turkish LLMs
Merhaba! 👋
Great leaderboard for Turkish LLMs! For Turkish enterprise and startup teams, cost per token often determines whether a Turkish-specific model or a frontier model like GPT-4o makes more business sense.
I built an open source LLM Evaluation Framework that adds these dimensions:
→ 💰 Cost per 1K tokens — Turkish model vs GPT-4o-mini vs Gemini Flash side by side
→ ⚡ Latency — real API latency, including from Turkey/EU servers
→ 🔍 Hallucination Rate — 0.0-1.0 score, critical for Turkish legal/news applications
→ 🎯 Accuracy — MMLU + TruthfulQA + custom benchmark support for Turkish tasks
→ 🧠 Reasoning Quality — chain-of-thought depth
Live demo (no API key): https://huggingface.co/spaces/vigneshwar234/llm-eval-demo
GitHub: https://github.com/vignesh2027/LLM-Evaluation-Framework
Happy to discuss adding Turkish-specific benchmark support!