Gemini 4 Krypton, Grok 4.8 and Qwen 4: What Is Confirmed Before You Budget
Updated November 12, 2026 · first published November 12, 2026
As of November 12, 2026, none of Gemini 4 Krypton, Grok 4.8 or Qwen 4 has a published price, model ID or release date that we could find from its vendor. Krypton comes from tester posts, Grok 4.8 from Elon Musk statements relayed by news sites, and Qwen 4 from a single conference remark that it is in training. Budget with published prices and treat these as scenarios.
Status at a glance
| Model | What the vendor has said | Price, ID, date |
|---|---|---|
| Gemini 4 Krypton | Nothing found from Google. Reported only through tester posts on X from about November 9, described as a first checkpoint. | None published |
| Grok 4.8 | Reported as announced by Elon Musk on X in September as a 2.5 trillion parameter model. A Japanese report says its reinforcement-learning stage was announced on November 10. We found no xAI model card. | None published |
| Qwen 4 | Alibaba said at its Apsara conference on September 22, 2026 that Qwen 4 is in training, per an NYU Shanghai summary of the announcement. | None published |
Most of this is secondhand. We did not find an official page for any of the three, so the table reflects what reputable-looking secondary sources say, not vendor documentation.
Where the rumours conflict
- Krypton. Sources describing the tests (OfficeChai) say it is not available in the Gemini app, AI Studio, Vertex AI or the Gemini API. Claims that it writes about twice the code in about twice the time as Gemini 4 Carbon come from side-by-side tests posted on X and are unverified.
- Grok 4.8. One aggregator page describes Grok 4.8 and a "Fast" variant as already available in some tools. Other sources say the model is still in post-training. We could not verify the availability claim against xAI, so we do not rely on it.
- Qwen 4. Several sites list Qwen 4 Max, Flash, Plus and 27B variants. NYU Shanghai's summary says Alibaba's published materials do not name them. A different model, Qwen3.8-Flash-Next, was released as a preview of the architecture and is not Qwen 4.
What is confirmed and priced instead
Google's own Gemini API pricing page lists gemini-4-carbon and gemini-4-argon at $2.00 per million input tokens and $10.00 per million output tokens (thinking included) through November 30, 2026, rising to $4.00 and $20.00 from December 1, 2026. That dated step is the one price change in this group that you can put in a forecast today. We read the part of the page available to us and found no Krypton entry. Google's Carbon model card is dated November 10, 2026 and gives a context window of up to 1 million tokens.
How to budget around unreleased models
- Forecast on published rates only. Store each rate with its effective date, as in our Q4 2026 price calendar.
- Make unreleased models a scenario, not a line item. If you model one, label it hypothetical and state the price you assumed. Do not copy a leaked figure into a finance document.
- Decide what would trigger a re-test. For example: a published price, a model ID in the API, and a measured result on your own workload. A leaderboard score alone is not a cost result.
- Watch the vendor page, not the video. Roundups of leaks mix confirmed announcements with unconfirmed ones. The GPT-6.2 case is covered in GPT-6.2: release date and pricing, where the same discipline applies.
What we have not verified
We have not confirmed any release date, model ID, price, benchmark or open-weight status for Krypton, Grok 4.8 or Qwen 4. Everything above is dated November 12, 2026 and will change as the vendors publish. For the model that is released, see Gemini 4 Carbon vs Claude Fable 5.2 and the December 1 price doubling.
Frequently asked questions
Is Gemini 4 Krypton released?
Not as far as we could find. As of November 12, 2026 we found no Google announcement, model ID or price. It has been reported only through tester posts from about November 9, and sources say it is not in the Gemini API.
Has Grok 4.8 been released?
We found no xAI model card, price or API documentation. Secondary reports say it was announced as a 2.5 trillion parameter model and is in reinforcement learning, with one aggregator claiming otherwise that we could not verify.
When will Qwen 4 be released?
Unknown. Alibaba said at its September 22, 2026 Apsara conference that Qwen 4 is in training, and per a summary of that announcement gave no release date or variant names.
What should an AI budget assume about unreleased models?
Use published prices with effective dates for available models, such as the December 1 increase on Gemini 4 Carbon and Argon, and treat unreleased models as clearly labelled hypothetical scenarios.
Related
- AI API price calendar: Q4 2026
- GPT-6.2: release date and pricing
- Gemini 4 Carbon vs Claude Fable 5.2
- Gemini 4 Argon price doubles on December 1
Want this applied to your stack? Bring the provider bills, gateway logs, and top workflows; we will map the cost drivers and savings path. Book a free audit →