Breaking

xAI Ships Grok 4.5, Touting Cost Efficiency at Half the Price

July 10, 2026 at 19:01 EDT

Foundation Models
Software Dev & Coding
AI Agents

xAI released its flagship model Grok 4.5 to the public on July 8, and Elon Musk claimed it delivers the best real-world ROI. Rather than topping raw benchmarks, the strategy foregrounds advantages in price and token efficiency.

July 8 · xAI

Grok 4.5 skips the benchmark crown — and bets everything on ROI

Elon Musk pitches xAI's new flagship as "Opus-class, but faster, more token-efficient and lower cost." Built on the 1.5-trillion-parameter V9 model, it wins no single benchmark — yet undercuts rivals sharply on cost per real-world task.

$2 / $6

Per million tokens (input / output)

4.2×

More token-efficient than Opus 4.8 max on SWE-Bench Pro

500K

Context window (tokens) · ~80 TPS throughput

Token efficiency: average output per SWE-Bench Pro task

Fewer tokens = lower cost per task. Lower is better.

15,954

Grok 4.5

67,020

Opus 4.8 max

≈4.2× fewer tokens to finish the same task

Raw benchmarks: close, but never first

DeepSWE 1.0Grok 62.0% · lead 66.1%

SWE-Bench ProGrok 64.7% · lead 80.4%

Terminal Bench 2.1Grok 83.3% · lead 84.3%

AA Intelligence Index4th (Elo 1543) · #1 on cost/task

The case for

Sits on the Pareto frontier of performance vs. cost. Fast, low token use, strong for long agentic tasks, code generation, and debugging — boosted by Cursor session training.

The caution

Around 4th on raw benchmarks and wins none outright — "strong for specific workloads, but not a universal winner."

Continue reading

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