Breaking

SpaceXAI Releases Coding-Focused 'Grok 4.5' at API Prices Below Opus

July 8, 2026 at 21:59 EDT

Foundation Models
Software Dev & Coding
AI Agents

On July 8, 2026, SpaceXAI released Grok 4.5, a flagship model built for coding and agentic tasks. It offers a 500k-token context window and prices its API at a fraction of Claude Opus, while claiming to surpass rivals on some benchmarks.

July 8, 2026 · xAI

Grok 4.5 chases "Opus-class" coding at a fraction of the price

A flagship model for code and agentic tasks with a 500k-token context window — priced far below Claude Opus and claiming to beat rivals on some benchmarks.

$2 / $6

Input / output per 1M tokens

500k

Token context window

4.2×

Fewer output tokens vs Opus on SWE-Bench Pro

~80

Tokens per second

Input price per 1M tokens — the main lever

Grok 4.5 undercuts Opus 4.8 by roughly 2.5–4× on token pricing.

$2.00

Grok 4.5

$5–8

Claude Opus 4.8

Benchmark scores — a mixed picture

Terminal Bench 2.1Grok leads

Grok 4.5

83.3%

Opus 4.8

78.9%

SWE-Bench Pro (resolve rate)Opus leads

Grok 4.5

64.7%

Opus 4.8

69.2%

DeepSWE 1.0Grok leads

Grok 4.5

62.0%

Opus 4.8

55.75%

Why developers are drawn in

Opus-class coding, faster and cheaper
High token efficiency for long-running tasks
Terminal-centric agentic workflows

Open questions

Trails rivals on SWE-Bench Pro & some DeepSWE metrics
Stability still being validated in public preview
Leaks suggest supplementary Cursor-data training

Built on a reported 1.5-trillion-parameter foundation, Grok 4.5 stakes its case against the Opus line on two levers:

Price

A fra

Continue reading

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