Grok 4.5: xAI’s Fast, Low-Cost Coding and Agent Model
2 min readxAI has released Grok 4.5, which it calls its smartest model to date, built specifically for coding, agentic tasks, and knowledge work. The launch pushes Elon Musk’s AI company deeper into the developer market, and it leans hard on two claims that matter to anyone paying for AI: speed and cost.
What Grok 4.5 Brings
Grok 4.5 was trained across tens of thousands of NVIDIA GB300 GPUs, with a heavy emphasis on reinforcement learning over hundreds of thousands of software-engineering tasks. xAI says the model was trained alongside Cursor, the popular AI coding tool, and it now serves as the default model in Grok Build, the company’s agentic coding environment.
Beyond code, xAI is pitching Grok 4.5 as an office-work engine. The company showed the model building multi-sheet Excel financial models, assembling PowerPoint decks using native shapes, and drafting documents in Word.
The Benchmark Picture
On xAI’s published benchmarks, Grok 4.5 is competitive but not a clean sweep. It leads on SWE Marathon, a long-horizon software task, with a 29 percent resolution rate ahead of Anthropic’s Opus 4.8 and Fable 5. On other coding tests such as DeepSWE and SWE Bench Pro, it trails the top scorers, which xAI lists as Anthropic’s Fable 5 and OpenAI’s GPT-5.5. In short, Grok 4.5 sits firmly in the frontier conversation without dominating it.
Speed and Price Are the Real Pitch
Where Grok 4.5 stands out is efficiency. xAI serves it at roughly 80 tokens per second and says it uses about 4.2 times fewer output tokens than Opus 4.8 to solve the same benchmark tasks. Pricing lands at $2 per million input tokens and $6 per million output tokens, undercutting several rivals. The company’s framing is blunt: the highest intelligence per unit of time and cost.
Why It Matters
The model race is shifting from raw benchmark bragging rights toward the economics of running agents at scale. A model fast and cheap enough to power long, multi-step coding jobs changes what developers can afford to automate. Grok 4.5 is available now in Grok Build, in Cursor on all plans, and through the xAI API, though EU access is still pending. Watch whether the efficiency claims hold up under independent testing.