German Firm TNG Unveils DeepSeek-TNG R1T2 Chimera: A Faster, More Efficient Open-Source LLM

TNG Technology Consulting GmbH, a German firm, has released DeepSeek-TNG R1T2 Chimera, a new large language model (LLM) built upon the open-source DeepSeek-R1-0528. Utilizing their innovative Assembly-of-Experts (AoE) method, R1T2 boasts significant improvements in speed and efficiency, achieving over 200% faster inference than R1-0528 while retaining over 90% of its reasoning capabilities. The model's concise outputs translate to lower compute costs. Released under the permissive MIT license and available on Hugging Face, R1T2 offers a cost-effective and efficient AI solution for enterprises and researchers.