While it may not reach the level of Fable or GPT 5.5, it is said to have advantages in cost reduction and speed when such a high-level model is not required.
Original Summary
SpaceXAI has released Grok 4.5, its latest model specialized in coding, agent tasks, and knowledge work. Trained on datasets of code, science, engineering, and mathematics, this model demonstrates outstanding performance in real engineering tasks and also ranked first in legal agent benchmarks. Notably, it possesses capabilities ranging from complex coding tasks such as Rust and C/C++ to building entire apps with a single prompt. With a fast speed of 80 TPS and high token efficiency, Grok 4.5 delivers intelligent results faster and more affordably.
▶ Original Source: https://x.ai/news/grok-4-5
