Google unveils Gemini 4 Carbon, topping benchmarks with lowest hallucination rate
Google released Gemini 4 Carbon, its latest AI model, which topped the Artificial Analysis Intelligence Index V5 benchmark, scoring 59—one point ahead of Claude Opus 5.5 and two points above OpenAI’s GPT-6 Astra. It also led in four of nine evaluations and offered lower cost-per-task pricing than competitors like Claude Opus 5.5 and Sonnet 5.5, though still more expensive than GPT-5 Astra or…
Key points
- Gemini 4 Carbon scores **59** on Intelligence Index V5, surpassing Claude Opus 5.5 by **1 point** and GPT-6 Astra by **2 points**
- Model claims **54** on AA-Omniscience v1.1, the highest hallucination score among 2026 frontier models on a **-100 to +100** scale
Google is rolling out Gemini 4 Carbon cautiously, prioritizing cybersecurity applications through its Fairwind program for cyber defenders before broader availability. Google’s controlled release mirrors Anthropic’s approach with Mythos in 2025, emphasizing safety and government collaboration for pre-release testing. While Gemini 4 Carbon excels in general intelligence and reliability, it trails Claude Code in programming tasks and Grok 4.7, Muse Wonder, and Kimi K3.1 Pro in knowledge-based work.
Model pages: Gemini 4 Carbon → · Gemini 4 Argon → · Gemini 3.8 Flash →
The story so far
4 episodes →- Google unveils Gemini 4 Carbon, topping benchmarks with lowest hallucination ratethis story
Google reveals new Gemini 4 Carbon AI model, narrowly beating Claude Opus 5.5 — frontier AI model is also the least likely to hallucinate, benchmarks claim
Tom's Hardware · 3 November 2026
Loading the full article…
This text was published by Tom's Hardware and written by Jon Martindale. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Generative AI & Models
All →- A new way of measuring AI models suggests China's threat to OpenAI and Anthropic may have been overblown · 1 src
- Anthropic and OpenAI may be testing Fable 5.5 and GPT-6.2 Astra · 2 src
- Anthropic may have a new top-tier model in the works, report says · 2 src
- Anthropic reportedly prioritizes safety over image generation for Claude · 2 src
- LongCat debuts LongCat-2.5 with 1.6T parameters and visual token support · 1 src
Comments
via GitHub Discussions