Breaking

Musk Claims Grok 4.5 Matches Claude Opus 4.7 in Capability

July 8, 2026 at 13:47 EDT

Foundation Models
Software Dev & Coding
AI Agents

On July 8, 2026, Elon Musk said xAI's new model Grok 4.5 is roughly on par with Anthropic's top model Claude Opus 4.7 in capability, while being faster and cheaper. Musk framed this as an internal assessment, stressing that xAI is pursuing real-world usefulness rather than benchmark scores. He highlighted that engineers at Tesla and SpaceX consider the model "genuinely useful."

July 8, 2026 · xAI

Musk claims Grok 4.5 rivals Claude Opus 4.7 — "faster and cheaper"

xAI says its new model matches Anthropic's top system on capability while running quicker and at lower cost. But there is a catch: the claim is a vendor self-report — no prices, no benchmarks, no third-party checks.

1.5T

V9 foundation model parameters

256k

token context window (Cursor build, unconfirmed)

independent benchmarks released

A 3× scale-up in one generation

Parameter count from the previous V8 base to the new V9 base.

~500B

V8 base

1.5T

V9 base (Grok 4.5)

What's actually disclosed

Training

V9 base + supplemental training on Cursor data, plus ongoing reinforcement learning.

Availability

Private beta for SpaceX & Tesla; public rollout expected around July 9.

Price & scores

Not disclosed — no per-token pricing, no SWE-bench Verified figures.

THE PITCH

Opus-class capability, but faster and cheaper
Tesla & SpaceX engineers call it "genuinely useful"
Suited to long-running agentic coding in Cursor
Part of a plan for a new base model every month in H2 2026

THE SKEPTICISM

"A self-report with zero benchmarks"
Internal engineer praise lacks third-party verification
Measuring against Opus already concedes a competitive edge
No public prices, no independent evaluations yet

Continue reading

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