Grok 4.5 is real, but you can’t use it yet. Elon Musk confirmed on June 28, 2026 that xAI’s newest model is already running in private beta with teams at SpaceX and Tesla, built on a fresh V9 foundation with roughly 1.5 trillion parameters. What he didn’t give is the one thing everyone wants: a public Grok 4.5 release date.
That gap between announcement and availability is where the confusion starts, especially with Grok 4.3 still the model most people are actually running and a much larger Grok 5 looming behind it. This article lays out exactly what xAI has confirmed about Grok 4.5, where it sits in xAI’s fast-moving lineup, the realistic timeline, how it compares to the rest of the Grok family, and how you can use Grok today while the public launch sorts itself out.
The Key Takeaways
- Grok 4.5 was announced June 28, 2026 by Elon Musk, but it is in private beta only with SpaceX and Tesla, with no public release date set.
- It runs on xAI’s new V9 foundation with ~1.5 trillion parameters, roughly 3x the V8-small architecture behind earlier Grok 4 builds.
- Training was supplemented with Cursor data to sharpen coding and technical performance.
- xAI’s internal evaluations claim Grok 4.5 is comparable to or better than Claude Opus, but no public benchmarks exist yet.
- It was originally targeted for late May 2026, so the launch is already running about a month late.
- The model you can use right now is Grok 4.3, with the much larger ~6 trillion-parameter Grok 5 still in training.
When Will Grok 4.5 Be Released?
As of June 29, 2026, Grok 4.5 has been announced but not publicly released. Musk revealed on xAI’s own channels that the model is in private beta, and xAI has not given a public launch date or a broader rollout plan. The model was originally slated for late May, which means it is already running roughly a month behind schedule.
That timing matters because xAI has been shipping at a relentless, roughly monthly cadence through 2026. A private beta limited to SpaceX and Tesla usually signals real-world stress testing inside Musk’s own companies before a wider release, which has been xAI’s playbook before. Based on that pattern, a public rollout measured in weeks rather than months is the most likely read, though xAI has a habit of slipping its own targets.
It’s worth setting expectations on that last point. xAI has repeatedly missed launch windows, with Grok 5 originally targeted for late 2025, then Q1 2026, then Q2 2026, all of which passed without a release. We’ll update this page with the confirmed Grok 4.5 date the moment xAI announces it.
Where Grok 4.5 Fits in xAI’s Lineup
To understand Grok 4.5, it helps to see how fast xAI has been iterating. In under three years the company has gone from a beta novelty bolted onto X to a frontier lab shipping new flagships every few weeks. Grok 4.5 is the latest rung on that ladder, sitting just above Grok 4.3 and just below the still-unreleased Grok 5.
| Version | Released | Headline change |
|---|---|---|
| Grok-1 | Nov 2023 | First beta, X Premium only |
| Grok-2 | Aug 2024 | Better reasoning, tool use, vision |
| Grok-3 | Feb 2025 | Large jump in reasoning |
| Grok-4 | Jul 2025 | Flagship, native tool-calling, real-time search |
| Grok 4.1 | Nov 2025 | 65% hallucination reduction |
| Grok 4.20 | Feb 2026 | Multi-agent system, 2M context, lower pricing |
| Grok 4.3 | Apr 2026 | Current public flagship, faster reasoning |
| Grok 4.5 | Jun 2026 | V9 foundation, ~1.5T params, coding focus (private beta) |
Read down that list and the pattern is clear. The point-releases between major numbers are where xAI tightens accuracy, speed, and cost, and Grok 4.5 looks like the biggest of those steps yet because it swaps in a new foundation model rather than just tuning the old one.
What Is Grok 4.5?
Grok 4.5 is xAI’s interim flagship, the step between the current public model, Grok 4.3, and the much larger Grok 5 still in training. It is the first model built on xAI’s new V9 foundation, and the headline change is scale plus a sharper focus on coding.
The V9 Foundation and 1.5 Trillion Parameters
Grok 4.5 runs on the V9 foundation with roughly 1.5 trillion parameters, which xAI says is about three times the size of the V8-small architecture that powered earlier Grok 4 variants. Parameters are a rough proxy for how much a model can learn and store, so a jump this large usually points to stronger reasoning and broader knowledge, even if raw size never guarantees a better model on its own.
For context, 1.5 trillion parameters puts Grok 4.5 in the same broad weight class as the frontier models from OpenAI, Anthropic, and Google, while still sitting well below the ~6 trillion parameters planned for Grok 5. xAI has not yet confirmed the context window or pricing for Grok 4.5, two specs that will shape how it competes once it’s public. Both are worth watching, since Grok 4.20 already pushed to a 2 million token context and aggressively low pricing.
Built for Coding
The most concrete detail is the training data. xAI supplemented Grok 4.5’s pipeline with data from Cursor, the popular AI coding environment, specifically to improve coding and technical competence. That is a deliberate shot at the part of the market where Claude and GPT have led, and it tells you where xAI thinks the next competitive battle is.
Coding has become the headline benchmark for frontier models because it is measurable, valuable, and hard to fake. Training on real-world data from a tool developers actually use, rather than generic code scraped from the web, is the kind of move that can close a gap quickly. It also explains why xAI’s early performance comparisons all point straight at coding-heavy rivals like Claude Opus.
Grok 4.5 vs Grok 4.3 vs Grok 5
It’s easy to lose track of which Grok is which right now, so here’s how the three relevant models line up. Grok 4.3 is what you can actually use today; Grok 4.5 is the near-term upgrade in beta; Grok 5 is the future flagship still being trained on xAI’s Colossus 2 supercomputer.
| Model | Status | Foundation / Parameters | Context window | Best for |
|---|---|---|---|---|
| Grok 4.3 | Public (Apr 30, 2026) | Reasoning-first flagship | 1M tokens | Available now, all-round use, native video input |
| Grok 4.5 | Private beta (Jun 28, 2026) | V9 / ~1.5T | Not yet confirmed | Coding and technical tasks (Cursor-trained) |
| Grok 5 | In training | Mixture-of-Experts / ~6T | ~1.5M expected | Future flagship, not yet released |
For the bigger picture on the 6-trillion-parameter model behind all of this, see our full breakdown of Grok 5. That model uses a Mixture-of-Experts design, where only a slice of its parameters fire on any single query, so it won’t burn 6 trillion parameters of compute on every response. xAI is reportedly training seven models at once on Colossus 2, including a 10-trillion-parameter variant, with full Grok 5 API access now looking like a Q3 2026 story.
What Grok 4.3 Can Do Right Now
Because Grok 4.5 is locked behind a private beta, Grok 4.3 is the model that actually matters for your work today. It launched April 30, 2026 as xAI’s most intelligent and fastest model to date, and it is more capable than many people realize.
Grok 4.3 ships with a 1 million token context window, always-on reasoning folded into its base behavior, and native video input that handles clips up to five minutes at 1080p. It can also produce finished files directly, generating PDFs, PowerPoint-style slide decks, and spreadsheets as outputs rather than just describing them. On top of that, xAI bundled a fast voice-cloning suite with the 4.3 release.
Pricing is aggressive at $1.25 per million input tokens and $2.50 per million output tokens, with prompt caching available at $0.20 per million for repeated context. In other words, you are not missing out by using Grok 4.3 while you wait. It is a genuinely strong model, and whatever Grok 4.5 adds will build on this foundation.
How Good Is Grok 4.5?
This is where caution matters. xAI’s internal evaluations suggest Grok 4.5 is comparable to, or better than, Anthropic’s Claude Opus. That’s a bold claim, and right now it is exactly that, a claim, because no independent or public benchmarks exist yet.
Vendor-reported numbers almost always flatter the home team, and a private beta limited to two friendly companies isn’t the same as open testing on LMArena or SWE-bench. The Cursor-trained coding focus makes the Opus comparison logical, since that’s Anthropic’s home turf, but treat any “beats Claude Opus” headline as unverified until third-party results land.
xAI does have a track record of strong showings on public leaderboards, with Grok 4.1 topping LMArena at launch and Grok 4.3 posting one of the lowest hallucination rates among frontier models. So the claim is plausible, not absurd. We’ll add real, independently tested benchmarks here once the model is public.
What to Expect When Grok 4.5 Goes Public
xAI hasn’t confirmed pricing or access details, but its recent history gives a reasonable template. Past releases have rolled out first to SuperGrok subscribers and on X, then to the xAI API, often within days of the public announcement. Expect Grok 4.5 to follow a similar staged path once the beta wraps.
On price, the trend has been downward, with Grok 4.20 launching at roughly 60% lower cost than the model before it. If that pattern holds, Grok 4.5 could land near or below Grok 4.3’s rates while offering a stronger model, which would make it one of the better value frontier options. The open questions are the context window and whether the coding gains show up in real developer workflows rather than just internal charts.
Is GROK 4.5 a Crypto Coin?
One quick disambiguation, because search results are messy here. There is a cryptocurrency token trading under names like “GROK 4.5” on some exchanges, and it has no connection to xAI’s AI model. xAI has not launched a coin, and Grok 4.5 the model is not a tradable asset. If you’re researching the AI, ignore the price charts.
How to Use Grok Right Now
Since Grok 4.5 isn’t public, your real options today center on Grok 4.3. There are three practical routes, depending on whether you want a chat app, raw API access, or every model in one place.
SuperGrok and the X app
The most direct route is a SuperGrok subscription, which gives you the latest Grok model inside the Grok app and on X. For current plans and token costs, check our Grok pricing guide before you commit to a tier, since xAI changes plans often.
The xAI API
If you’re a developer, the xAI API gives you Grok 4.3 at $1.25 input and $2.50 output per million tokens, with the 1 million token context window and prompt caching to keep costs down on repeated context. This is the path most likely to get Grok 4.5 the moment it ships.
Every model in one Mac app
If you’d rather not lock into a single model, Fello AI runs Grok alongside Claude, ChatGPT, and Gemini in one native Mac and iOS app. That matters here because when Grok 4.5 opens up, you can put its “beats Opus” claim to the test against the real Claude Opus 4.8 in the same window, with the same prompt, and judge for yourself. To see how Grok stacks up against the rest of the field, browse our regularly updated ranking of the best AI models.
Conclusion
Grok 4.5 is confirmed, capable on paper, and stuck in private beta with no public Grok 4.5 release date yet. The smart move is to watch for the public launch, treat the Claude Opus comparison as unproven for now, and keep using Grok 4.3, which is already a strong, well-priced model. Bookmark this page; we’ll update it the moment xAI sets a date and the first independent benchmarks appear.
FAQ
When is Grok 4.5 coming out?
There is no public release date yet. As of June 29, 2026, Grok 4.5 is in private beta with SpaceX and Tesla teams. It was originally targeted for late May 2026, so it is running about a month behind, and a wider rollout within weeks is the likely next step.
How many parameters does Grok 4.5 have?
Grok 4.5 is built on xAI’s V9 foundation with roughly 1.5 trillion parameters, which xAI says is about three times the size of the V8-small architecture used in earlier Grok 4 variants.
Is Grok 4.5 better than Claude Opus?
xAI’s internal evaluations claim Grok 4.5 is comparable to or better than Claude Opus, but no independent or public benchmarks exist yet. Treat the claim as unverified until third-party testing is available.
What’s the difference between Grok 4.5 and Grok 5?
Grok 4.5 (~1.5 trillion parameters) is an interim model shipping now in beta. Grok 5 is the larger ~6 trillion-parameter Mixture-of-Experts flagship still in training, expected around Q3 2026. Grok 4.5 arrives first.
Can I use Grok 4.5 right now?
Not yet, unless you’re on a SpaceX or Tesla beta team. The public model you can use today is Grok 4.3, available through a SuperGrok subscription, the xAI API, or a multi-model app like Fello AI.
