Capturing data…
Capturing data…
FRI, 30 OCT · 88 ITEMS
Google Blog, Hacker News, Engadget + 6 more · AI & Technology
9 outletsFraming varies: Coverage is largely aligned on the announcement, but some sources (e.g., Ars Technica) emphasize that the model cannot be used by the public, while others (e.g., Engadget, The Verge) highlight the performance claims; The Register offers a more skeptical tone with 'diamond in the rough'.
Google officially announced Gemini 4 Carbon, a new frontier large language model that is a refined version of its Gemini 4 Argon model. The model claims to achieve top scores on AI benchmarks, particularly in complex reasoning and scientific tasks, and is priced at $2 per million input tokens and $10 per million output tokens. The announcement comes with benchmark comparisons and availability details.
The claim is directly from Google's official blog, a primary source, making the existence and pricing authoritative. The performance claims are self-reported and should be independently verified, though third-party benchmarks (Artificial Analysis, BenchAlign) show similar high rankings. Overall, the core facts of the model launch are solid.
Genuinely new as of October 2026, announced less than 24 hours ago according to the search results.