Grok 4.8: Release Date, What to Expect, How It Compares
7 min read
Last updated October 31, 2026. How we cover news: we write from the official announcement and docs (linked below) and point you to the tools on Unloc. Not sponsored.
Short answer: Grok 4.8 has not been released as of 31 Oct 2026, and xAI has announced no release date. The newest Grok model xAI has published is Grok 4.7, which reached the API on 21 September 2026. Everything specific about Grok 4.8, including that it is a 2.5-trillion-parameter model, comes from posts by Elon Musk rather than from an xAI announcement. On 30 October Musk said "it's taken longer than expected".
Sponsored
Your tool here — sponsor this list
Fixed $149/month on Draft with AI. A subscription — not Claim #1.
Pin this category · $149/moStatus: Not released as of 31 Oct 2026. Last verified: 31 Oct 2026 (Dubai time), against xAI's Grok 4.7 announcement, the xAI API release notes, xAI's news page and Musk's public posts on X. xAI's site now carries the SpaceXSI name; we say xAI here.
What's confirmed
- Official: the latest Grok model is Grok 4.7, published 21 Sep 2026 (xAI's Grok 4.7 announcement; xAI API release notes). It follows Grok 4.6 (12 Aug 2026) and Grok 4.5 (8 Jul 2026), both listed in the same release notes.
- Official (by absence): xAI's news page and API release notes, checked 31 Oct 2026, contain no Grok 4.8 or Grok 5 entry, no date and no specification.
- Claimed by Musk, not xAI: Grok 4.8 is a 2.5T model, it was due to finish training in the week of 14 Sep 2026, and it "will be a noticeable improvement" over Grok 4.7. Details and quotes are in the table below.
- Not known: release date, price, context window, benchmark scores, and whether the final name will be 4.8.
How we label claims: Official means published by xAI or SpaceX on their own sites or in their filings. Claimed by Musk means something Elon Musk wrote or said, which is not an xAI announcement. Reported means a third-party outlet or aggregator said it. Rumour means unsourced or leaked. We quote Musk's exact words and never present them as xAI facts.
What has been said about Grok 4.8, with dates
| Date | Claim | Label | Source |
|---|---|---|---|
| 20 Apr 2026 | Replying to a user's joke ladder of future versions, Musk wrote "4.6 → 3T / 4.7 → 6T / 4.8 → 10T / 4.9 → ??? / 5.0 → AGI", ending with a shrug and a laughing emoji. Informal, and the sizes did not match his later statement. | Claimed by Musk | X post |
| 4 Aug 2026 | On SpaceX's Q2 call: "Grok 4.5 was a huge improvement. We have Grok 4.6 coming out probably next week. Then, Grok 4.7 is about three or four weeks from today." | Claimed by Musk (official transcript) | SpaceX Q2 2026 earnings transcript |
| 14 Sep 2026 | "Grok 4.8, which is a 2.5T model trained with our new C++ software stack, will finish training this week and start RL." | Claimed by Musk | X post |
| 14 Sep 2026 | "Grok 4.7 should be roughly on par with Opus 5.0, not 5.1. Better in some ways, worse in others. We need to fix multimodal performance. Grok 4.8 will be a noticeable improvement. Grok 4.9 is probably Astra/Fable class. Grok 5 maybe better than anything. We shall see." | Claimed by Musk | X post |
| 7 Oct 2026 | "most @Bot requests are pretty simple and will be handled by a lightning-fast version of Grok 4.8 when that comes out." "@Bot" appears to mean Grok Bot; this is about how simple requests would be routed. | Claimed by Musk | X post |
| 30 Oct 2026 | "I know it's taken longer than expected, but Grok 4.8 will not be a disappointment. We're not optimizing for benchmarks and I really don't care how Grok does on those, but it will be incredibly useful." | Claimed by Musk | X post |
Two things stand out. First, the size has moved: the April ladder put 4.8 at 10T, and the September post says 2.5T. Second, in none of these posts is there a date. "Finish training this week and start RL" on 14 September describes the next stage of training, not a launch.
What xAI itself has published
xAI has published nothing about Grok 4.8. Its latest announcement is Grok 4.7, which it calls its most capable model for coding and knowledge work, built on "a new, larger base model compared to Grok 4.6" with a longer reinforcement learning run. The xAI post gives no parameter count for it, and we found no parameter count in xAI's own Grok pages, so every trillion-parameter figure you see is a Musk statement or a report of one. For what a launch would look like, the xAI API release notes show the pattern: xAI lists a model, its API name, context window and prices on the day it ships.
The earlier Grok 4.7 timing is a useful reference, with a caveat. On 4 August Musk said Grok 4.6 was "probably next week" and Grok 4.7 "three or four weeks from today". The release notes show Grok 4.6 on 12 Aug, eight days later, and Grok 4.7 on 21 Sep, 48 days later. The gaps between xAI's last three releases were 35 days (4.5 to 4.6) and 40 days (4.6 to 4.7); Grok 4.7 is now 40 days old. That is history, not a schedule, and no release date for Grok 4.8 has been announced.
SpaceX's third-quarter results came out on 20 October. The transcript we read (Motley Fool's SpaceX Q3 call transcript, a secondary source) has executives saying they are accelerating the Grok release cadence, and it does not name Grok 4.8 or give a date for it.
How Grok 4.8 could compare: what is public today
There are no Grok 4.8 benchmarks, so any comparison would be invented. What can be compared is the bar a launch will be judged against. In its Grok 4.7 post, xAI published its own results against models it picked itself. These are xAI's numbers, not independent tests:
| Benchmark (xAI's table) | Grok 4.7 | Grok 4.6 | GPT-5.6 Sol | Fable 5.1 |
|---|---|---|---|---|
| CursorBench 4.0 | 46.3% | 40.4% | 41.7% | 51.8% |
| DeepSWE v1.1 | 71.0% (high effort) | 65.2% | 72.7% | 70.0% |
| Terminal-Bench 4.0 | 37.6% | 20.3% | 37.3% | 57.9% |
Source: xAI's Grok 4.7 announcement, Official, 21 Sep 2026. xAI compared Grok 4.7 with GPT-5.6 Sol rather than the newer GPT-6 line. Since then the field has moved on: Anthropic shipped Opus 5.5 on 22 Sep, OpenAI shipped GPT-6.1 Terra on 26 Oct, and Google announced Gemini 4 Carbon on 28 Oct. A Grok 4.8 launch would be measured against those, not against Grok 4.7.
Musk's own framing is modest on benchmarks. On 14 September he said Grok 4.7 was "roughly on par with Opus 5.0, not 5.1", with multimodal performance still to fix, and on 30 October he said he doesn't "really care" how Grok does on benchmarks. That matters when you read leaks: a claim that Grok 4.8 beats a named rival is a Rumour until xAI publishes results or independent testers do.
Where Grok 4.8 sits relative to Grok 5
Musk's 14 September post lists Grok 4.8, then Grok 4.9, then Grok 5. In a separate reply the same day, asked how close Grok 4.8 gets to AGI, he wrote "That will be Grok 5". We cover that model, and why its dates keep moving, in Grok 5: release date, size claims and what we know.
What to do while you wait
- Use what is shipped. Grok 4.7 is available now, and the API name per xAI's release notes is
grok-4.7. Check the release notes before building around any "4.8" claim. - Don't wait on a rumour. If a tool or site says Grok 4.8 is live, check xAI's news page first. Some sites have published false "launched" stories about Grok models.
- Compare options. See best Grok alternatives, Grok vs ChatGPT, Grok vs Claude and Grok vs Gemini. For agent-style work, see Grok Bot and the main Grok page.
FAQ
When is Grok 4.8 coming out?
No release date has been announced. As of 31 Oct 2026 xAI's news page and API release notes list no Grok 4.8. Musk said on 30 October that it has "taken longer than expected" without giving a date. Treat any specific date as a guess.
Is Grok 4.8 a 2.5 trillion parameter model?
Musk said so on 14 September 2026: "Grok 4.8, which is a 2.5T model trained with our new C++ software stack". xAI has not published a size. Musk's own April ladder gave a different figure (10T), so treat the number as his claim, not a specification.
Will Grok 4.8 beat Claude and GPT?
Nobody outside xAI has published results. Musk said Grok 4.8 "will be a noticeable improvement" over Grok 4.7, which he described as roughly on par with Opus 5.0. That is a claim about a model nobody can test yet, not a benchmark.
Will Grok 4.8 come before Grok 5?
Musk's 14 September post lists them in that order, with Grok 4.9 in between. On the SpaceX Q2 call on 4 August he said Grok 5 "should be out before the end of this year". Neither is an announced date. See our Grok 5 page.
Can I use Grok 4.8 today?
No. The newest model you can use is Grok 4.7, shown in xAI's release notes as grok-4.7. The "lightning-fast version of Grok 4.8" Musk mentioned on 7 October for "@Bot" requests was tied to the model coming out, so it is not available either.
Submit a tool if it belongs on the board. Listings are reviewed before they go live.
Share your learnings on socials
Other resources
View all
Xero vs QuickBooks: which accounting software is better?
Pick Xero for cloud accounting with no per-user licence fees, a big app marketplace and the JAX AI assistant. Pick QuickBooks for AI agents across the product, a desktop option and plans that scale to complex inventory and projects.
Mistral Large 4 open weights: what Le Chonk is and who can use it
Mistral released the open weights of Mistral Large 4 (Le Chonk) on October 29, 2026, calling it the best open-weight model for coding. Here's what the trillion-parameter model does, how it compares with Claude and other open models, and how you can actually use it.
Windsurf vs GitHub Copilot: which AI coding tool is better?
Pick Windsurf, now Devin Desktop, for a standalone AI editor with Tab completions and a command centre for many agents and models. Pick GitHub Copilot for AI inside the IDE you already use and across GitHub, from code review to cloud agents.