Gemini Carbon: HLE-Diamond Debut?

On October 9, 2026, Business Insider reported that Google employees were internally testing a Gemini 4 model codenamed "Carbon," separate from the checkpoint later released as Gemini 4 Argon (https://www.businessinsider.com/google-employees-test-new-gemini-4-model-argon-barium-carbon-2026-10). This market will resolve to "Yes" if the Google "Carbon" model added to the Humanity's Last Exam results at https://agi.safe.ai/ has an HLE-Diamond Accuracy of at least the specified percentage at 12:00 PM ET on the calendar date following the date on which it first appears on the site. Otherwise, this market will resolve to "No". A qualifying model must be a Google AI model named "Carbon" (including variants or successor branding such as Gemini Carbon, Gemini 4 Carbon, or similar naming) or be confirmed to be the same model referenced in the report by Google or by a consensus of credible reporting. A model listed under any other name, including a new Gemini model (e.g., Gemini 4.1) or an update to Gemini 4 Argon, will qualify only if it is confirmed by June 30, 2027, 11:59 PM ET to be the same model previously codenamed "Carbon". Models merely derived from, distilled from, or incorporating improvements from Carbon will not qualify unless confirmed to be the same model. The percentage displayed as "HLE-Diamond Accuracy" for the model in its result card on the "AI Progress on HLE-Diamond" chart at https://agi.safe.ai/ will be used to resolve this market. This market will resolve solely based on the HLE-Diamond Accuracy, regardless of the model's Calibration Error, chart position, other configuration scores, any note that a score was computed on a sample of questions, or any underlying granular or unrounded data presented elsewhere. If the Center for AI Safety (CAIS), which maintains the official Humanity's Last Exam results, publishes a qualifying score (defined below) for the model, and the model has not been displayed on the chart by 11:59 PM ET on the third calendar day (ET) after the first such publication, that score will be used instead, rounded to one decimal place. If multiple qualifying scores for the model have been published by that time, the highest will be used. A qualifying score is one issued under CAIS's name (for example, on https://agi.safe.ai/, https://lastexam.ai/, or https://dashboard.safe.ai/, including jointly with Scale AI) that is clearly identified as overall accuracy on the full HLE-Diamond question set (1,000 questions, released September 22, 2026), without tools, at the reasoning setting labeled "high" in HLE-Diamond's main results. No score other than the chart value or a qualifying score will count, even if no qualifying score is published. This includes scores on the original Humanity's Last Exam or HLE-Rolling, "reasoning max" or "with tools" results, Reasoning or Knowledge partition scores, values available only through an API, data file, or page source, scores from Scale AI's separate leaderboards, other evaluators, or the model's developer, and any score where there is reasonable doubt that it meets the conditions above. A model first appears on the site on the earlier of the date (ET) it is first displayed on the chart and the date (ET) a qualifying score for it is first published. If the chart value is used, the check described in the first paragraph takes place at 12:00 PM ET on the calendar date following the model's first display on the chart; if a qualifying score is used, it takes place at the end of the three-day period described above. If multiple qualifying models are added to the Humanity's Last Exam results on the same calendar date (ET), the model with the highest HLE-Diamond Accuracy will be used for resolution. Models added to the results on the calendar date following the initial qualifying model's first appearance will not be considered. If a model is first added to the Humanity's Last Exam results but is subsequently removed and not re-added, such that no HLE-Diamond Accuracy …