Gemini 4 Carbon, day 3: Perplexity calls it the top research model as cadence debate grows

Gemini 4 Carbon arrived on October 28, two weeks after Gemini 4 Argon, and immediately took the #1 spot on Artificial Analysis Intelligence Index v5.0 with a score of 59, up from Argon's 50. Sundar Pichai highlighted cyber numbers: 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploitation, and 71.0% on CWE-bench for patching. Google says the model leads 15 of 17 benchmarks, including a 96.2% F1 on GraphWalks BFS.

The new angle today comes from customers. Perplexity CEO Aravind Srinivas shared early evals calling Carbon the strongest model his team has seen for long-form research, with a new state of the art on the Wide-And-Deep-Research (WANDR) benchmark. He said it will come to Perplexity Pro and Max users inside Computer once enough token capacity is available.

Google built Carbon for multi-day autonomous development workflows and proactive cyber defense. Because of its exploitation scores, access to its frontier agentic cyber capabilities starts with a vetted tier of defensive organizations rather than open API access. That puts Google in the same gated posture OpenAI took with its Daybreak cyber models.

The reaction is split. Commentary argues the release cadence matters more to builders than a one-point lead, and some developers say they cannot keep benchmarks current. Security practitioners complain the gating leaves smaller blue teams behind. Leaks of a 'Gemini 4 Xenon' already circulate on Reddit, suggesting the next drop may come soon.