xAI has launched Grok 4.5, its most advanced model to date, specifically designed for programming, agentic tasks, and knowledge work. Trained with Cursor on NVIDIA GB300 GPUs, Grok 4.5 shows strong performance in coding benchmarks, achieving 62% on DeepSWE 1.0 and excelling in the SWE Marathon test. The model prioritizes speed and efficiency, processing at 80 tokens per second and using significantly fewer tokens than Opus 4.8, which is reflected in its competitive pricing. Additionally, Grok 4.5 enhances office productivity by generating complex Excel models, PowerPoint diagrams, and Word documents, and serves as the default model for the Grok Build tool. Availability is currently limited outside the EU, with wider access anticipated in mid-July. AI
IMPACT Sets new SOTA on SWE-Marathon benchmark and offers competitive pricing, potentially impacting developer tool adoption.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
Read on Mastodon — fosstodon.org →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
