Sunday, July 26, 2026 · Beirut, Lebanon Newsletter · About · Contact
Artificial Intelligence

xAI ships Grok 4.5, an ‘Opus-class’ model tuned for coding and agents

xAI released Grok 4.5 on July 8, built for coding and agentic work, trained alongside Cursor and priced at $2 and $6 per million tokens.

I
IT Magazine Staff July 10, 2026 · 2 min read
Artificial Intelligence

xAI has released Grok 4.5, positioning its newest model squarely at software developers and autonomous agents rather than general chat. The company published the model on 8 July 2026 and began rolling it out to the public the following day, describing it as its most capable release so far.

Grok 4.5 was trained on data spanning coding, science, engineering and mathematics, with the company saying it emphasised both stronger reasoning and greater efficiency. A distinguishing detail is that xAI folded in developer session data from the coding tool Cursor during supplemental training, which the firm credits for the model’s performance on real engineering tasks. Training ran in xAI’s Memphis data centres on tens of thousands of NVIDIA GB300 GPUs.

Elon Musk framed the launch in competitive terms, calling Grok 4.5 an “Opus-class model, but faster, more token-efficient and lower cost.” xAI is pricing the model at $2 per million input tokens and $6 per million output tokens, and has made it available to developers through its API and via the AI SDK for direct integration.

The pitch reflects a broader shift among frontier labs toward models optimised for agentic workflows, where systems chain together tool calls, write and edit code, and complete multi-step tasks with limited human oversight. By training against real coding sessions and stressing token efficiency, xAI is betting that developers will weigh running costs as heavily as raw benchmark scores when they choose a model for production use. The release also lands in an unusually crowded stretch of the calendar, with several rival labs shipping updates in the same window.

For teams in the region evaluating coding assistants and agent platforms, Grok 4.5’s lower per-token pricing and API-first availability make it a candidate worth benchmarking against incumbents on their own workloads before committing.

Sources: xAI, TechCrunch

Drafted with AI assistance and reviewed by the IT Magazine news desk. Facts verified against the linked sources. Spotted an error? See our Corrections Policy.

Related stories