Grok 4.5 Launches Across Platforms, Tops Coding Benchmarks with High Token Efficiency

Grok 4.5 has been officially released across Web, X, iOS, and Android, marking Grok's evolution from a simple chatbot into a comprehensive work platform. The model tops multiple coding benchmarks and acts as a new efficiency benchmark for agentic workflows due to its extremely low token consumption.

Confirmed

Platform and Ecosystem Expansion: According to posts by @XFreeze and @zetalyrae, Grok 4.5 is now available on Grok Build, Cursor, and API. New capabilities include automation triggered by emails or schedules, orchestration of up to a thousand agents, and the ability to handle entire projects directly.

Coding and Efficiency Performance: @XFreeze noted that Grok 4.5 achieved high scores in both mean reward and binary pass on the Long-Horizon Terminal-Bench. Information reposted by @elonmusk highlights that Grok Build can save up to 5x tokens on coding tasks, costing only about 35 cents per task, placing it in an optimal quadrant on the efficiency frontier.

Why it matters

Grok 4.5 significantly improves processing efficiency and reduces operational costs for agentic tasks. Developer @haider1 praised its high cost-effectiveness but recommended using it as an execution worker paired with GPT-5.6 Sol for architectural planning and review, offering a highly efficient workflow reference for complex software engineering.

2026-07-26 ~ 2026-07-27 · 5 related posts

Full story(20 episodes)→

Primary sources