Google is preparing to ship a new AI model called Gemini 3.8 Flash as soon as Wednesday, September 2, 2026, according to a report from The Wall Street Journal. The model is reportedly known inside Google as “Skimaki,” and its main selling point is a fix for the one area where Gemini has consistently trailed OpenAI and Anthropic: coding. Alphabet’s stock moved on the news, and the report has already been picked up by financial trackers including Stocktwits and TradingView, along with outlets such as Investing.com and CryptoBriefing.
This is a fast-moving story built on reporting rather than an official Google announcement, so the details below separate what has been reported from what remains unconfirmed. What’s clear is that Google is under pressure to close a coding gap, and the market is already pricing in that effort before a single line of a changelog has gone public.
Google Prepares to Ship Gemini 3.8 Flash This Week, WSJ Reports
The Wall Street Journal’s report says Alphabet’s Google is readying Gemini 3.8 Flash for release as soon as Wednesday, citing people familiar with the matter. Alphabet Inc. trades under the ticker symbols GOOGL and GOOG on the Nasdaq, and it is the parent company of Google. The Journal frames the model as an attempt to close a coding gap that has dogged Gemini for much of 2026, an area where rivals OpenAI and Anthropic have generally been seen as ahead.
Secondary coverage has moved quickly. Investing.com ran a piece headlined around Google narrowing its AI coding gap, and financial-market aggregators including Stocktwits and TradingView picked up the same reporting within hours. CryptoBriefing’s writeup adds a helpful data point: it places Gemini 3.7 Flash’s launch on August 13, 2026, which means a 3.8 Flash reveal on September 2 would arrive roughly three weeks later. That is an unusually tight turnaround even by the standards of a company that has been shipping Flash-tier updates at a brisk pace this year, a cadence we tracked when Business Insider first reported employees testing a 3.8 Flash preview back on August 27.
None of this is a formal Google product announcement. There is no public changelog, no benchmark page, and no pricing sheet for a model called Gemini 3.8 Flash as of this writing. What exists is a Wall Street Journal report, a codename, and a wave of financial and tech coverage repeating it. That distinction matters for anyone trying to plan around this release, whether they’re a developer deciding which model to build against or an investor reading into the stock move.
Inside the Codename: What “Skimaki” Tells Us About Google’s Priorities
Internal codenames rarely leak with this much consistency across outlets, and “Skimaki” has now shown up in reporting from The Wall Street Journal down through smaller aggregators covering the same wire. Google routinely assigns internal names to models before they get a public marketing label, and the pattern here echoes the way Gemini 3.7 Flash and earlier Flash-tier models moved from internal testing to general availability over the past year.
What’s notable is the framing attached to the name. Every report so far ties Skimaki directly to coding performance rather than to multimodal features, longer context windows, or consumer-facing tools. That’s a shift in emphasis. Earlier in 2026, Google’s Gemini messaging leaned heavily on things like the Gemini 3.5 Transcribe rollout inside Gmail, which was about accessibility and productivity, not software development. Skimaki, by contrast, is being pitched almost entirely as an engineering tool, aimed squarely at developers and enterprise coding workflows.
How Google Tested 3.8 Flash Against Anthropic’s Claude Opus
According to the Journal, Google ran head-to-head internal tests inside Jetski, its internal coding platform, comparing Skimaki against Anthropic’s Claude Opus model. Google employees reportedly preferred the new model in those coding-focused evaluations. That’s a meaningful detail because Opus has been treated internally at plenty of engineering shops, including Google’s own, as one of the stronger options for complex coding tasks over the past year.
CryptoBriefing’s sourcing adds texture here too, describing early tester feedback as calling Skimaki “more of a refinement than a revolution,” with the bulk of the improvement centered on cutting down verbose, overly long outputs, a complaint that had followed earlier Flash releases. That’s a more modest claim than “beats Opus across the board,” and it’s worth holding onto that nuance. Preferring a model in an internal side-by-side test on a specific coding platform is not the same as a published benchmark showing a clear win across standardized coding suites. No such benchmark has been made public for Skimaki as of September 2.
Anthropic, for its part, has kept up its own release pace this year. We’ve covered reports of Anthropic prepping two additional Claude models, so the Skimaki comparison lands in the middle of an active back-and-forth between the two labs on coding-specific capability, not a static target.
Why Coding Became Google’s Weakest Link in the AI Race
Coding has become one of the highest-value battlegrounds in enterprise AI in 2026, because it’s the use case with the clearest, most measurable return: fewer engineering hours, faster ticket resolution, more automated pull requests. OpenAI and Anthropic built early reputations in this space with coding-oriented tooling and agentic features that let a model plan, write, and revise code with less hand-holding. Google’s Gemini line, by comparison, has more often been discussed for multimodal strengths, search integration, and consumer products than for agentic coding.
That gap matters commercially. Enterprise customers evaluating AI vendors for developer tooling tend to run their own internal bake-offs, not unlike the Jetski tests described in the Journal’s report. If Google’s own engineers are choosing Skimaki over Opus internally, that’s a signal Google wants external customers to hear, even before the model has cleared a public release. It’s also a signal of urgency: shipping three weeks after 3.7 Flash suggests Google is willing to accept a faster, rougher release cadence in exchange for staying visible in the coding conversation rather than waiting for a more polished, less frequent upgrade cycle.
GOOGL Stock Reaction: What the After-Hours Numbers Show
Alphabet’s stock moved on the report, though the exact magnitude depends on which outlet’s snapshot you’re reading. One report tracking the news put GOOGL up around 0.7% in after-hours trading, after the stock had ended the regular session nearly 1.3% lower. A separate report describes Alphabet shares gaining 0.6% in after-hours trading on the Tuesday session tied to the same news cycle. Both figures point the same direction: a modest positive after-hours bump tied specifically to the Gemini 3.8 Flash reporting, layered on top of a down day in the broader session.
That’s a fairly typical pattern for AI-model news that arrives after a report rather than an official launch event: enough to move sentiment among traders watching the story in real time, not enough to represent a re-rating of the stock. The more interesting question is why a coding-focused Flash-tier model, historically the cheaper, faster, less flagship end of Google’s model lineup, is moving the needle on Alphabet’s share price at all.
| Metric | Reported Figure | Reporting Context |
|---|---|---|
| Regular session close (day of report) | Down ~1.3% | Same-day report tracking the WSJ story |
| After-hours move (same report) | Up ~0.7% | Reaction after Gemini 3.8 Flash news broke |
| After-hours move (separate report) | Up ~0.6% | Tuesday session, same news cycle |
| Ticker symbols | GOOGL / GOOG | Alphabet Inc., Nasdaq |
| Model tier affected | Flash (not Pro or Ultra) | Lower-cost, higher-volume tier of Gemini |
The answer likely has less to do with Flash itself and more to do with what it represents: proof that Google can move quickly on the exact capability, coding, where investors have been most skeptical of Gemini’s competitive position. A three-week turnaround from 3.7 to 3.8 Flash is Google’s way of showing its release engine still runs fast, even if the underlying model is described by early testers as an incremental refinement rather than a leap.
Alphabet’s Rocky 2026: The Backdrop to This Rally
This isn’t happening in a vacuum. Reuters reported in July 2026 that Alphabet shares had fallen more than 10% since the end of April, a stretch that coincided with concerns over delays in Gemini’s roadmap and some executive turnover at the company. Against that backdrop, a report suggesting Google can ship a targeted, coding-focused model on a three-week cycle reads as a small but useful counter-narrative for a stock that had been under pressure for months on execution concerns rather than demand concerns.
It also lines up with a broader theme we’ve tracked across Alphabet’s business this year, including Google Cloud’s growth outpacing AWS and Azure in recent quarters. Coding-capable models feed directly into cloud consumption: every enterprise customer that adopts Gemini for software development is also a customer running more inference workload through Google Cloud and Vertex AI. A stronger coding story for Gemini isn’t just a model story, it’s a cloud-revenue story.
Competitive Landscape: Google, OpenAI and Anthropic on Coding
The coding-model race now effectively has three anchor players, each approaching it from a different angle.
Anthropic: Opus as the Internal Benchmark
Anthropic’s Claude Opus is explicitly named in the Journal’s report as the model Google’s own engineers were comparing Skimaki against inside Jetski. That alone tells you where Anthropic sits in the pecking order: it’s the model other labs benchmark themselves against internally, even before those comparisons go public. Anthropic has continued to iterate on its lineup through 2026, including the reported work on additional Claude models we covered separately.
OpenAI: The Incumbent Pressure
OpenAI is named alongside Anthropic in the reporting as one of the two labs Google is trying to catch up to on coding capability. OpenAI has spent much of 2026 retiring and consolidating older model lines, a cycle we detailed when OpenAI retired DALL-E, GPT, and o3 models against a set of internal deadlines, while pushing newer models toward agentic and developer-focused workflows.
Google: Speed as a Strategy
Google’s apparent strategy, based on the reporting so far, is less about leapfrogging on raw capability and more about compressing its release cycle so it never falls too far behind for too long. A three-week gap between Flash releases, even if each individual release is incremental, keeps Google in the headlines and in the internal bake-offs that enterprise buyers run before choosing a coding assistant vendor.
Table: Reported Timeline of the Gemini 3.8 Flash Story
| Date | Development | Source |
|---|---|---|
| August 13, 2026 | Gemini 3.7 Flash reaches general availability | CryptoBriefing |
| August 27, 2026 | Report of internal staff testing a “Gemini 3.8 Flash Preview” on Jetski | Business Insider |
| Late August 2026 | Production deployment testing continues throughout the month | CryptoBriefing |
| September 1-2, 2026 | Report that release, codenamed “Skimaki,” could land as soon as Wednesday | The Wall Street Journal |
| September 1-2, 2026 | Coverage spreads to financial and tech outlets; GOOGL moves after-hours | Investing.com, CoinDesk, Stocktwits, TradingView |
Laid out this way, the pattern is less “surprise announcement” and more “accelerating drip of leaks,” each one adding a bit more specificity, from an unnamed preview, to a codename, to a specific day of the week for release.
What a Three-Week Turnaround Says About Google’s Release Strategy
CryptoBriefing’s reporting also references comments attributed to Google CEO Sundar Pichai about targeting a monthly release cadence for Flash-tier models going forward. If accurate, that would formalize what has, until now, looked like an accelerating but informal pattern. A monthly Flash cadence would put Google on a materially faster public release schedule than either OpenAI or Anthropic have committed to for their flagship coding-capable models.
There’s a tradeoff embedded in that strategy. Faster releases mean less time for public benchmarking, less time for third-party red-teaming, and, based on early tester feedback described in the reporting, a real risk that individual releases feel more like patches than milestones. Google appears to be betting that staying visible and iterating quickly outweighs the risk of shipping a Flash update that testers describe as a refinement rather than a breakthrough.
Historical Context: From a Slow Start to a Monthly Cadence
It’s worth remembering how far Google’s release posture has shifted over the past two years. Google’s early Gemini rollout drew criticism for moving slower than OpenAI’s iteration pace and for uneven early results on coding tasks specifically. The 2026 version of Google looks almost unrecognizable next to that starting point: multiple Flash releases in a single year, an internal testing platform (Jetski) purpose-built for rapid coding evaluation, and now a reported push toward monthly public releases.
That shift didn’t happen by accident. It tracks with Alphabet’s broader pattern this year of trying to prove execution speed across the business, not just in Gemini. Faster Flash releases are Google’s most visible, lowest-cost way of demonstrating that speed to a market that had grown skeptical after the roadmap delays covered by Reuters over the summer.
Market Impact: Enterprise Coding Tools and Vertex AI
For enterprise buyers, the practical impact of a coding-focused Flash release lands on procurement decisions already in motion. Companies evaluating AI coding assistants for their engineering teams tend to run pilot programs across two or three vendors before standardizing. A Flash-tier model that internal Google testers preferred over Claude Opus, even informally, gives sales teams a fresh talking point in those pilot conversations, regardless of whether that preference holds up once outside developers get their hands on it.
There’s also a pricing angle. CryptoBriefing’s reporting cites Flash-tier pricing running through year-end 2026, positioning the tier as Google’s low-cost, high-volume option rather than a premium flagship product. That pricing strategy matters for the coding market specifically, where usage volumes (think: an AI assistant running against every pull request in a large codebase) can get expensive fast on a per-token basis. A cheaper, “good enough” coding model at Flash pricing could pull budget-conscious engineering teams away from pricier flagship alternatives, even if it doesn’t top every benchmark.
What Prediction Markets and Analysts Are Watching
Interestingly, the uncertainty around exactly when Gemini 3.8 Flash ships has spilled into betting markets. A question on Manifold Markets asking when Google will release a Flash-tier model at version 3.8 or higher, open to the public without invitation, was pricing a 53% probability of release before September 11, rising to roughly 71% before October 1 and 94% before October 21. That spread suggests bettors see the Wednesday timeline as plausible but far from locked in, with meaningful probability mass still sitting weeks past the reported date.
That kind of skepticism is reasonable given the pattern in this story so far: an August 27 preview report became a September 1-2 “as soon as Wednesday” report, and neither has been confirmed by an official Google release yet as of this writing. Traders and engineering leads alike are effectively pricing in the gap between “reported” and “shipped.”
Risks and Caveats: Why “As Soon As Wednesday” Isn’t Guaranteed
A few caveats are worth repeating plainly. First, “as soon as Wednesday” is reported timing based on people familiar with the matter, not a Google press release or blog post. Release dates built on that kind of sourcing slip regularly, sometimes by days, sometimes by weeks, without any official comment from the company involved. Second, the internal preference for Skimaki over Claude Opus comes from Google’s own staff on Google’s own internal tool, which is not the same as an independent, reproducible benchmark. Third, no pricing, context-window specification, or capability list for Gemini 3.8 Flash has been published by Google itself as of September 2, 2026.
None of that means the reporting is wrong. The Wall Street Journal, Investing.com, CoinDesk, and CryptoBriefing are converging on a consistent set of details: the name, the codename, the testing platform, and the general timing. But converging reports are still reports, and readers making real decisions, whether that’s an engineering team picking a coding assistant or an investor reading into the stock move, should treat Wednesday as a strong possibility rather than a confirmed date.
Predictions: What Comes Next for Gemini and GOOGL
- Expect Google to confirm or formally launch Gemini 3.8 Flash within days of this report, even if the exact date slips slightly past Wednesday, given how consistent the leak trail has been across multiple outlets.
- If Pichai’s reported monthly-cadence comments hold, expect a Gemini 3.9 or a follow-up Flash update to surface within four to six weeks of whatever date 3.8 Flash actually ships.
- Anthropic and OpenAI are likely to respond with their own coding-focused updates or marketing pushes rather than let an internal “engineers preferred it over Opus” claim sit unanswered in the press.
- Expect enterprise sales conversations at Google Cloud and Vertex AI to lean on the coding-improvement narrative in Q4 2026, tying model releases more directly to cloud consumption pitches.
- GOOGL’s reaction to the actual launch, once it’s official, is likely to be smaller than the after-hours bump on the report itself, since markets have already partially priced in the news ahead of confirmation.
What This Means for Developers Right Now
For developers and engineering leads, the practical advice is to wait for the official release before making tooling decisions. Early tester feedback describing the model as “more of a refinement than a revolution” is a reasonable expectation to carry into any hands-on testing once the model is publicly accessible. Teams already running Gemini 3.7 Flash in production coding workflows should watch for an official migration path and updated pricing before assuming Flash-tier costs stay flat at 3.8. Teams evaluating Anthropic’s Claude Opus or OpenAI’s coding-focused offerings alongside Gemini should treat Google’s internal preference claim as a data point worth testing themselves, not as a settled comparison.
Frequently Asked Questions
What is Gemini 3.8 Flash?
Gemini 3.8 Flash is a reported new AI model from Google, internally codenamed “Skimaki,” described in a Wall Street Journal report as focused on improving coding capability compared to earlier Gemini models.
When will Gemini 3.8 Flash be released?
The Wall Street Journal reports it could be released as soon as Wednesday, September 2, 2026, though this is based on people familiar with the matter rather than an official Google announcement, and the exact date could shift.
Why is it called “Skimaki”?
Skimaki is the internal codename reported for the model before its public release. Google, like most large AI labs, commonly uses internal codenames during development and testing that differ from the eventual public product name.
How did GOOGL stock react to the news?
Reports describe GOOGL moving modestly higher in after-hours trading, with one report citing a roughly 0.7% after-hours gain following a regular session that closed about 1.3% lower, and a separate report citing a 0.6% after-hours gain on the same news cycle.
Was Gemini 3.8 Flash tested against Claude Opus?
Yes. The Wall Street Journal reports that Google ran internal head-to-head tests inside its Jetski coding platform, comparing the new model against Anthropic’s Claude Opus, and that Google engineers reportedly preferred the new model in those coding-focused tests.
What is Jetski?
Jetski is described in reporting as Google’s internal coding platform, used by Google staff to test and evaluate AI models on coding tasks before public release.
Is this an official Google announcement?
No. As of this writing, this is based on reporting from The Wall Street Journal and secondary coverage from outlets including Investing.com, CoinDesk, and CryptoBriefing. Google has not published an official changelog, benchmark, or pricing page for a model called Gemini 3.8 Flash.
How does this compare to the previous Gemini 3.8 Flash preview report?
An earlier report from Business Insider on August 27, 2026 described Google staff testing an internal “Gemini 3.8 Flash Preview” build. This Wall Street Journal report adds new specifics: the “Skimaki” codename, a possible Wednesday release date, and the market’s reaction in GOOGL shares.
