Breaking

xAI Ships Grok 4.5 as OpenAI Rolls Out GPT-5.6

July 8, 2026 at 18:46 EDT

Foundation Models
Software Dev & Coding
AI Agents

xAI (now operating as SpaceXAI) and OpenAI both put new flagship models built around coding and agentic work into public hands on July 8-9, 2026, releasing Grok 4.5 and GPT-5.6 in quick succession.

July 8–9, 2026 · Frontier Model Race

Two Flagship Models Ship in 48 Hours

xAI's Grok 4.5 and OpenAI's GPT-5.6 land back to back — signaling a shift from benchmark chasing toward real-world coding, agentic workflows, speed and cost efficiency.

1.5T

V9 foundation params — ~3× prior size

Intelligence Index — #4 overall, +16 vs 4.3

Coding Agent Index — on par with GPT-5.5

80t/s

Grok 4.5 output speed

Output-token efficiency on SWE Bench Pro

Grok 4.5 uses ~4.2× fewer output tokens than Opus 4.8 for the same coding work — the core cost-performance edge.

4.2×

Opus 4.8
tokens

1×

Grok 4.5
tokens

Grok 4.5 — reported specs & benchmarks

Context window

500k tokens (down from 1M)

Pricing (in/out)

$2 / $6 per M tokens

Terminal Bench

83.3%

SWE Bench Pro

64.7%

τ³-Banking (agentic)

33% (near top)

Input

Text + vision

▲ What developers praise

Low latency preserves coding "flow"
Solves in seconds what took rivals minutes
Cursor & tool integrations seen as a moat
Refined via SpaceX / Tesla beta use

▼ What critics flag

"Opus-class" claims need independent verification
Context window cut from 1M to 500k
EU rollout delayed to mid-July
Long inputs >200k double the token cost

Continue reading

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