Share Facebook Twitter LinkedIn Pinterest Email The new model appears to be a fine-tuned version of Meta’s Llama-70b. beats Benchmarks GPT4o model Nvidias opensource