Elon Musk used a single X thread on July 25, 2026 to make two separate claims: that Grok 4.5 sits alone with Claude Opus 5 on the "Pareto frontier" of AI models, and that Grok 4.6 and 4.7 are already on the calendar — two and four weeks out, respectively. Neither claim came with a benchmark sheet. Here's what's actually verifiable, what's marketing, and what Grok 4.5 delivers right now while the timeline plays out.
TL;DR
| Question | Answer |
|---|---|
| When does Grok 4.6 ship? | Musk said "2 weeks" from July 25, 2026 — roughly August 8, 2026. No official xAI confirmation yet. |
| When does Grok 4.7 ship? | Musk said "4 weeks" from July 25, 2026 — roughly August 22, 2026. Same caveat. |
| Is this an official xAI roadmap? | No. It's a founder post on X, not a blog post, changelog, or press release from xAI. |
| What's the "Pareto frontier" claim about? | Musk says only Grok 4.5 and the newly shipped Claude Opus 5 (July 24, 2026) beat every other model on cost-vs-capability simultaneously. |
| What does Grok 4.5 do today? | 500K-token context at ~80 tok/s, $2/M input tokens, topped Ramp's 150K-invoice extraction test. |
| Should I wait for 4.6/4.7 before picking a model? | No — evaluate on Grok 4.5's shipped benchmarks; treat 4.6/4.7 as unverified until xAI publishes specs. |
What Musk Actually Posted
The sequence, in order, on July 25–26, 2026:
- "Grok 4.5 and Opus 5 are alone on Pareto frontier" — a claim that only these two models currently offer the best capability for their cost, with no chart or methodology attached.
- "Grok 4.6 in 2 weeks and Grok 4.7 in 4 weeks" — posted as a reply in the same thread, effectively an informal roadmap announcement.
Both posts crossed hundreds of thousands of engagements within hours, but as with prior xAI announcements — see the Grok 4.5 public launch coverage — the pattern holds: Musk announces on X first, and an official model card, benchmark suite, or Cursor/xAI blog post follows days to weeks later, if at all on the stated schedule.
Why the Opus 5 comparison matters
The timing isn't a coincidence. Claude Opus 5 shipped July 24, 2026 — one day before Musk's post. Framing Grok 4.5 as co-leading the frontier with Anthropic's just-released flagship is a direct competitive response, not an independent benchmark claim. Anthropic hasn't published a Pareto-frontier chart placing Opus 5 next to Grok 4.5, and xAI hasn't either — this is Musk's read of the market, not a peer-reviewed comparison.
For context on how Grok 4.5 fared against Anthropic's previous flagships, the Grok 4.5 vs Claude Opus 4.7/4.8 breakdown found Opus 4.8 still ahead on SWE-Bench Pro, with Grok competitive on cost and Snorkel GDPval+ professional-task scores.
What Grok 4.5 Delivers Right Now
Independent of the 4.6/4.7 timeline, Grok 4.5 — live since July 8–9, 2026 — has concrete, shipped numbers:
- 500,000-token context window at approximately 80 tokens/second
- $2 per million input tokens — among the more efficient prices at its capability tier
- Topped Ramp's internal invoice-extraction benchmark across 150,000 real business documents
- Powers Grok Build, xAI's coding tool with voice prompts and parallel agent execution
- Reported usage surges on Augment Code and OpenRouter following launch
These are the numbers to evaluate against today — not the unreleased 4.6 or 4.7.
What's Unverified
Treat the following as claims, not facts, until xAI or a third party confirms them:
- Exact ship dates for Grok 4.6 (~Aug 8) and Grok 4.7 (~Aug 22) — Musk's public track record on stated AI timelines has slipped before
- What changes between 4.5 → 4.6 → 4.7 — no word on whether these are new pretrained models, post-training updates, or pricing/context tweaks
- The Pareto-frontier claim itself — no chart, no benchmark suite, no third-party replication as of this writing
- Whether "Opus 5" in Musk's post refers to the same Claude Opus 5 Anthropic shipped July 24 — reasonable assumption given the timing, but not stated explicitly
The Bottom Line
Musk's July 25 post gives a two-model release cadence — Grok 4.6 in roughly two weeks, Grok 4.7 in four — layered on top of an unbenchmarked claim that Grok 4.5 and Claude Opus 5 are the only models on the efficiency frontier. Both are worth tracking, neither is verified yet. If you're choosing a model today, benchmark against what's shipped: Grok 4.5's $2/M pricing and 500K context, or Opus 5's published evals. Revisit this post once xAI ships an actual model card for 4.6 or 4.7.
Related on explainx.ai
- Grok 4.5 Public Launch — SpaceXAI, Benchmarks, Pricing
- Grok 4.5 vs Claude Opus 4.7 and 4.8 Comparison
- Claude Opus 5 Release Speculation — Timeline and Leaks
- Grok Build — Open Source SpaceXAI Coding Tool
- ARC-AGI-3 Opus 5 Leaderboard, July 2026
- AI Benchmark Claims Fact-Check 2026
- Musk vs Altman Feud History, July 2026
Release dates for Grok 4.6 and 4.7 reflect Elon Musk's public statements as of July 26, 2026 and are unconfirmed by xAI. Verify against xAI's official channels before making purchasing or migration decisions.