Google is about to roll out a new AI model. Employees say they're testing another that's way better.
Google employees are testing a newer version of the company's flagship AI model, Gemini 4. The latest internal release is codenamed Carbon.

Official score
Data as of · Model v31
Six axes weighted by the chosen profile. Valuation is scored half on absolute scales, half against the median of the company's industry peers — its sector when the industry is too thin. Measures that do not describe a company's economics are replaced by declared substitutes; none applies here. When financial strength is measured in the publication, a bounded deduction (at most 14 points) applies by company family: a standard grid (leverage, cash flow, dilution), REIT-specific thresholds, and regulated utilities are judged within their own lens. Where the required measure is not published — banks, insurers, holding companies — nothing is invented: the gap weighs on coverage, of which 15% is reserved for financial strength. Three axes — valuation, profitability and growth — are scored only when their measures cover at least 60% of the axis; below that, the axis is declared absent rather than estimated. A grade is published only if the profile's DOMINANT axis is scored and the profile's coverage reaches 60%. Entering a ranking additionally requires every axis weighted by the profile, the lens's essential signals and 70% overall coverage. Incomplete grades remain provisional and are not ranked. The conservative profile also requires its financial-strength data; holding companies are not part of general rankings. Same method for everyone, published and reproducible.
This describes the security's characteristics versus its sector — it is not a buy or sell recommendation. Polaris is not a registered adviser.
Third-party analyst consensus, for information. Technical read = descriptive price signals. Neither is Polaris advice.
Up to four releases and the next consensus, to understand differences from expectations.
Reserved for the Elite plan.
An integrated reading — technical, fundamental, macro, news feed and latest earnings — by a Polaris engine.
Diagnostic reading, not predictive. Candlestick patterns have a modest and inconsistent statistical edge — used here as triggers for thesis reflection, not as buy/sell signals.
Click an indicator to understand what it means.
| Period | Est. | Actual | Surprise |
|---|---|---|---|
| Q2 2026 | 2.98 $ | 2.85 $ | — |
| Q1 2026 | 2.71 $ | 2.62 $ | — |
| Q4 2025 | 2.71 $ | 2.82 $ | — |
| Q3 2025 | 2.40 $ | 3.10 $ | — |
Third-party analyst estimates, for information — not a Polaris recommendation.
Google employees are testing a newer version of the company's flagship AI model, Gemini 4. The latest internal release is codenamed Carbon.


How well entrenched is this company against its competitors? Graded against the median of its sector peers.
Trader members only — peer analysis
The grade relies on comparison with sector peers, included from the Trader plan up.
Official score
Data as of · Model v31
Six axes weighted by the chosen profile. Valuation is scored half on absolute scales, half against the median of the company's industry peers — its sector when the industry is too thin. Measures that do not describe a company's economics are replaced by declared substitutes; none applies here. When financial strength is measured in the publication, a bounded deduction (at most 14 points) applies by company family: a standard grid (leverage, cash flow, dilution), REIT-specific thresholds, and regulated utilities are judged within their own lens. Where the required measure is not published — banks, insurers, holding companies — nothing is invented: the gap weighs on coverage, of which 15% is reserved for financial strength. Three axes — valuation, profitability and growth — are scored only when their measures cover at least 60% of the axis; below that, the axis is declared absent rather than estimated. A grade is published only if the profile's DOMINANT axis is scored and the profile's coverage reaches 60%. Entering a ranking additionally requires every axis weighted by the profile, the lens's essential signals and 70% overall coverage. Incomplete grades remain provisional and are not ranked. The conservative profile also requires its financial-strength data; holding companies are not part of general rankings. Same method for everyone, published and reproducible.
This describes the security's characteristics versus its sector — it is not a buy or sell recommendation. Polaris is not a registered adviser.
Third-party analyst consensus, for information. Technical read = descriptive price signals. Neither is Polaris advice.
Up to four releases and the next consensus, to understand differences from expectations.
Reserved for the Elite plan.
The Terminal tier is coming — advanced tools and early access for the most engaged members.
Per-ticker discussion is reserved for paid plans
Upgrade to a paid plan (Essential and up) to read and join member discussions on this ticker.
An educational space for members — this is not investment advice.
An integrated reading — technical, fundamental, macro, news feed and latest earnings — by a Polaris engine.
Sign in to write your investment thesis on this stock. Polaris will bring it back when context shifts.
Emerging-growth profile (conviction + risk), by transparent criteria. Terminal-exclusive.
Sign in to access research dossiers.
Polaris analysis is coming soon — a synthetic read of the stock, grounded in the Cote and the data. Educational only, never advice.