SpaceXAI Launches New Generation Flagship Large Language Model Grok 4.5
SpaceXAI launches flagship model Grok 4.5, featuring a 1.5T parameter MoE architecture and co-trained with Cursor, with performance benchmarked against Claude Opus. The model's inference speed reaches 80 TPS, programming capabilities surpass Opus 4.8 on DeepSWE Bench, and Token efficiency improves by over 4 times. It supports 500k context, will upgrade to 1 million next week, and a 2T parameter version will be released by the end of the month.
SpaceXAI launches flagship model Grok 4.5, featuring a 1.5T parameter MoE architecture and co-trained with Cursor, with performance benchmarked against Claude Opus. The model's inference speed reaches 80 TPS, programming capabilities surpass Opus 4.8 on DeepSWE Bench, and Token efficiency improves by over 4 times. It supports 500k context, will upgrade to 1 million next week, and a 2T parameter version will be released by the end of the month.
This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.