Breaking

xAI Releases Coding-Focused Grok 4.5, Co-Trained With Cursor

July 9, 2026 at 16:15 EDT

Foundation Models
Software Dev & Coding
AI Agents

xAI (SpaceXAI) unveiled Grok 4.5 on July 8, 2026, a new model specialized for coding and agentic tasks, and made it available the same day. Co-trained with the AI-native IDE Cursor, it is the company's first developer-focused model and is positioned to build end-to-end functional apps from a single prompt.

July 8, 2026 · xAI

Grok 4.5 Ships for Coders — Half the Price, Built with Cursor

xAI's first developer-focused model, co-trained with the AI-native IDE Cursor, builds end-to-end apps from a single prompt — betting on speed, cost and real-world efficiency to challenge Anthropic and OpenAI.

4.2×

fewer output tokens than Opus 4.8 on SWE-bench Pro

80 TPS

generation speed, ~2× the token efficiency of rivals

500k

token context length

Output pricing per 1M tokens

Grok 4.5 undercuts flagship rivals — the core of its "intelligence per cost" pitch.

Grok 4.5

~$12+

Rival flagships

Input: $2 / 1M ($0.50 on cache hit)
"Roughly half the price of rivals."

Benchmark resolution rate — SWE Marathon

29.0%

Grok 4.5

26.0%

Claude Opus 4.8

Also reported to beat GPT 5.5 and Opus 4.8 on average pass rate in one third-party professional-task evaluation.

Where it wins

Polished Three.js sims, 3D games & dashboards from one prompt
High first-try success rate and fast generation
Beat Claude Opus 4.8 on frontend performance
Strong "intelligence per cost" edge

The caveats

Complex tasks still need extra prompting
Fable 5 leads on long-horizon & creative architecture
No EU support yet (expected mid-July)
Higher pricing on very long inputs

Available now

Shipped the same day via API, all Cursor plans and Grok Build — positioned as the practical, speed-and-cost model against Anthropic's flagship-class Claude Fable 5 and Opus 4.8.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.