Breaking

xAI Launches Grok 4.5 With Same-Day API Access, Aimed at Coding and Agents

July 8, 2026 at 15:40 EDT

Foundation Models
AI Agents
Software Dev & Coding

xAI released Grok 4.5 on July 8, 2026, positioning it for coding, agentic tasks and knowledge work, and made it available immediately through the xAI API. The company describes it as its "smartest model built for coding, agentic tasks, and knowledge work," and its website foregrounds Grok 4.5.

July 8, 2026 · xAI

Grok 4.5 arrives — betting on price, not just the podium

xAI's new flagship for coding, agentic tasks and knowledge work lands third on a leading coding benchmark — but does the work at roughly one-tenth the cost of the top model.

~11×

cheaper per coding task than the top-ranked model

3rd

overall on CursorBench, at 66.7% (vs 70.5% leader)

$2/$6

API price per 1M input / output tokens

Cost per task — the story to scale

CursorBench coding evaluation · lower is better

~$1.51

Grok 4.5

66.7% · 3rd

~$17.32

Fable 5 Max

70.5% · 1st

Nearly the leader's score for a fraction of the spend — 3.8 points behind, roughly 11× cheaper per task.

Early praise

Standout cost efficiency from day one
Faster than Opus, far faster than Fable
Improved writing & instruction-following
Real-time X integration and fast research

Open questions

No comprehensive benchmark table at launch
Long-context factual consistency has varied
Real-world agentic validation still ahead
Must prove differentiation vs Claude & OpenAI

The strategic shift

After SpaceX acquired xAI in February 2026, the focus moved from benchmark chasing toward agentic, coding and cost-efficient frontier performance.

Grok 4 · tool use & search → SpaceX acquisition → Grok 4.5 · agentic & coding

Deployed via the xAI API, Databricks, Amazon Bedrock and tool integrations (Cursor, Warp, Office add-ins). Configurable reasoning effort: low / medium / high.

Continue reading

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