Breaking

xAI Releases Grok 4.5 With Free Access in Coding Tools

July 10, 2026 at 05:16 EDT

Foundation Models
Software Dev & Coding
AI Agents

xAI launched Grok 4.5 on July 8, 2026, a new model tuned for coding and agentic work, and opened a limited-time free window in the coding tools Grok Build and Cursor. With a 500K-token context window, the release is pitched squarely at software developers. According to the official announcement, Grok 4.5 is the first model trained jointly with the code editor Cursor.

July 8, 2026 · xAI

Grok 4.5 Bets on Cost-per-Task in the Coding-Agent Race

A model co-trained with Cursor on real developer workflows, tuned for agentic coding with a 500K-token context — pitched on "intelligence per dollar and time" against pricier frontier rivals.

500K

token context window (down from 1M, restore planned)

Coding Agent Index — on par with GPT-5.5

~91

tokens/sec measured generation speed

Token efficiency — the core pitch

Roughly 2–4.2× fewer tokens on comparable tasks. Relative token use per task:

1×

Grok 4.5

up to 4.2×

Rival frontier models

Fewer tokens per agent loop → lower cost on sustained runs. SWE-Bench Pro avg output ≈ 15,954 tokens.

API pricing / 1M tokens

Input$2

Output$6

Cached input$0.50

Requests above 200K context may cost ~2×.

Benchmarks

Terminal-Bench 2.183.3%

SWE-Bench Pro64.7%

Developer praise

Handles large codebases in one 500K-context pass
Fewer tokens in agent loops on long runs
Clearly cheaper and faster than Opus 4.8
Multi-file refactors finished faster; easy to wire into Aider, Cline, Hermes

Caveats to verify

Costs rise beyond 200K context
EU availability slated for mid-month
Post-trial pricing and rate limits unclear
Test agentic-workflow stability on your own benchmarks

Continue reading

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