Running
Agents
2
Greek NLU Benchmark Leaderboard
🏆
Show a ranked leaderboard of Greek NLU benchmark scores
AI & ML interests
None defined yet.
Recent Activity
Organization Card
🛡️ Sovereign European AI ⚖️ Open Weights 🇬🇷 Built in Greece 🗣️ Greek · English MIT · CC-BY-4.0
Open models, datasets, and demos from KIEFER SA — Greece's AI infrastructure company building the next generation of language, speech, and agentic systems for Europe. We publish open-weight models, fine-tuning recipes, and evaluation benchmarks designed for production — from Apple Silicon edge deployment to enterprise inference on NVIDIA GPU clusters.
🧠 Open-Source Models
KIEFER publishes open-weight models that lead Greek-language AI — Sophea K1, a 28B LLM that tops GreekMMLU at 83.34%, and Sophea Canary ASR, a bilingual Greek + English speech model reaching 2.90% WER on Greek, with a native Apple Silicon MLX build. Every model ships with open weights, reproducible recipes, and permissive licensing (MIT / CC-BY-4.0) — browse them all in the Models tab below.
🌐 The Sophea Ecosystem
KIEFER SA's integrated suite of sovereign AI products — language models, speech intelligence, robotics, and energy infrastructure. Every product is built on open-weight Sophea models and served on KIEFER's sovereign European GPU infrastructure.
| Product | URL | Description |
|---|---|---|
| Sophea AI | sophea.ai | The core platform — LLM, ASR, TTS, RAG, and agentic workflows |
| Sophea Meet | meet.sophea.ai | AI meeting intelligence — real-time transcription, summaries, and action items powered by Sophea ASR + K1 |
| Sophea Nous | nous.sophea.ai | Advanced reasoning and knowledge platform built on Sophea K1 and GraphRAG |
| AI Robots | ai-robots.ai | Humanoid and quadruped robotics AI — perception, planning, and control powered by Sophea models |
| Hyperion-X | hyperion-x.ai | AI-driven energy infrastructure and BESS optimization — physical AI for sovereign energy systems |
⚡ Infrastructure
All Sophea models are developed and served on KIEFER's sovereign AI infrastructure — NVIDIA GPU clusters deployed in Greece, designed for data-residency compliance and low-latency European inference. Our roadmap targets 1,000+ GPU capacity to support the full Sophea model family at enterprise scale.
📎 Resources
- 🌐 Website — kiefer.gr
- 🦉 Try Sophea K1 — sophea.ai
- 📄 Licenses — MIT (Sophea K1) · CC-BY-4.0 (Sophea ASR models)
- 🤝 Enterprise & partnerships — info@kiefer.gr · info@sophea.ai
