🔍 Read the full analysis: Gemini 4 Carbon Is The New #1 AI Model, By One Point. Here Is What The Ranking Leaves Out on ThorstenMeyerAI.com
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Gemini 4 Carbon leads Artificial Analysis’ Intelligence Index v5 with 59 points, one point ahead of Claude Opus 5.5. The ranking reflects nine tests, not a universal measure of AI capability; Carbon leads on four and trails on five. Google says it is initially giving access to a limited group of cyber defenders, with no public release date announced.
Google’s Gemini 4 Carbon has moved into first place on the Artificial Analysis Intelligence Index v5, scoring 59 points, one point ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The narrow lead puts Carbon at the top of this particular nine-test benchmark, but Google has not opened the model to the public.
Artificial Analysis’ index gives Carbon 59 points, compared with 58 for Claude Opus 5.5 at its maximum reasoning setting, 57 for GPT-6 Astra at maximum, and 55 for GPT-6.1 Sol at maximum. Google’s earlier Gemini 4 Argon scores 50. Carbon ranks first on four of the nine tests and does not lead the other five.
Its strongest results include 69% on HLE-Diamond, one point above Claude Opus 5.5, and 65% on Terminal-Bench-Science, two points above GPT-6 Astra. On AutomationBench-AA, a test of business workflows across software tools, Carbon scores 78%, slightly ahead of Argon. It also leads AA-Omniscience, which rewards models for avoiding incorrect answers as well as answering correctly.
The model’s results are not uniformly strongest. Claude models lead on office-style knowledge work, Claude Opus 5.5 leads on the real-world coding test, GPT-6 Astra leads on a physics-research test, Kimi K3.1 Pro leads on terminal tasks, and Meta’s Muse Wonder Full leads on document reasoning, according to the source report. The reported cost is $7.76 per index task at standard prices: less than Claude Opus 5.5’s $10.04, but more than GPT-6 Astra’s $6.24 and substantially more than GPT-6.1 Sol’s $1.27.
Infographic 01 · What the #1 is made of
Number one overall. First in four of nine tests.
The Intelligence Index is a decathlon score: nine tests, one number. Carbon wins on reasoning, science and careful knowledge, and trails on office work and coding.
The nine tests behind the score
Carbon at its highest setting (high) against the best model on each test.
| Test | Weight | Carbon | Leader or runner-up |
|---|---|---|---|
| HLE-Diamond Hard reasoning + knowledge questions | Weight 10% | Carbon 69% | Carbon leads Next: Claude Opus 5.5, 68% |
| Terminal-Bench-Science Science research work in a terminal | Weight 15% | Carbon 65% | Carbon leads Next: GPT-6 Astra, 63% |
| AA-Omniscience v1.1 Knowledge, wrong answers penalized | Weight 15% | Carbon 54 | Carbon leads Next: Claude Opus 5.5, 48 |
| AutomationBench-AA Business workflows across apps | Weight 5% | Carbon 78% | Carbon leads Next: Gemini 4 Argon, 78% |
| AA-Briefcase Office-style knowledge work | Weight 15% | Carbon 1,609 Elo | Behind Claude Sonnet 5.5, 1,812 |
| FrontierCode 2.0 Code a maintainer would merge | Weight 15% | Carbon 41% | Behind Claude Opus 5.5, 46% |
| CritPt v2 (hard) Research-level physics | Weight 10% | Carbon 63% | Behind GPT-6 Astra, 71% |
| Terminal-Bench 4.0 Agentic terminal tasks | Weight 10% | Carbon 65% | Behind Kimi K3.1 Pro, 67% |
| GDP.pdf Reasoning over documents | Weight 5% | Carbon 28% | Behind Muse Wonder Full, 33% |
The surprise in the knowledge test
- Accuracy: Claude Fable 5.1 68%, Claude Opus 5.5 66%, GPT-6.1 Sol and GPT-6 Astra 62%, Carbon 58%
- Hallucination rate: Carbon 11%, Claude Opus 5.5 and GPT-6 Astra about 50%
- Carbon wins because it rarely bluffs, not because it knows the most
Read the weights
- Each of the four blocks counts 25%: knowledge work, coding, science, general
- Carbon trails in two of the four heaviest tests (15% each): office-style work and real-world code
- Hallucination rate = wrong answers among all questions not answered correctly
Source: Artificial Analysis, 28 Oct 2026; Index v5 announcement, 27 Oct 2026; Artificial Analysis model data, retrieved 29 Oct 2026. Scores rounded as in the article; Carbon tested at high, other models at their highest tested setting.
Powered by Thorsten Meyer AI
A One-Point Lead Has Limits
Carbon’s first-place ranking may draw attention from companies choosing models, but the one-point margin does not establish that it is best for every task. Its score combines several kinds of work, and the results show meaningful variation: a model can top the overall table while trailing competitors in coding, office work or document analysis.
The pricing figures also make the task-level results relevant to buyers. A lower price per benchmark task does not necessarily mean lower costs for a particular company: the source reports standardized index-task costs, not a bill for a customer’s own workload. Teams comparing models should test them on their actual work and compare cost per completed task, accuracy and reliability rather than relying on the leaderboard alone.
For everyday users, the immediate effect is limited because Carbon is not publicly available. If it later reaches general users, the source report suggests its advantages may be most apparent on long, difficult tasks rather than ordinary chat. That is an expectation, not a measured result for public users.
Products Worth Considering
Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Black
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How the Index Is Built
Artificial Analysis describes its Intelligence Index as a combined score from nine tests, grouped into four equally weighted categories: knowledge work, coding, science and general capability. The index version used for this ranking launched on October 27, 2026. Tests include agent work in terminals, business workflows, coding tasks and difficult knowledge questions.
The source report says 60% of the index uses private test sets or held-out answers, intended to reduce the chance that models have memorized the questions. It also says many tasks are lengthy and involve agents completing work rather than simply answering short prompts. Those features can make results more representative of some practical uses, while making runs more expensive.
The AA-Omniscience result illustrates how test design affects rankings. Carbon scores 58% on plain accuracy, below several named rivals, including Claude Fable 5.1 at 68% and Claude Opus 5.5 at 66%. But the test also penalizes wrong answers. The report gives Carbon an 11% hallucination rate, compared with about 50% for Claude Opus 5.5 and GPT-6 Astra, and says that restraint helps Carbon lead on the test’s scoring. That does not mean Carbon answers the most questions correctly.
Release Timing and Test Limits
Google has not provided a public release date in the supplied report. It says access will begin with a subset of cyber defenders, followed by trusted partners and then developers, businesses and consumers, but the timing and eligibility for those stages remain unspecified.
The source text is also incomplete: it ends during a sentence about Google and supplies no further details about that claim. The report does not provide a full methodology for every test, independent replication of the results, or enough information to determine how Carbon performs across different customer workloads. The index measures performance on its selected tests and settings; it is not a general verdict on which model is best.
Pricing is described as an introductory API rate of $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20 respectively. The supplied material does not specify when those later rates take effect or whether all users will receive the introductory price.
Access Rolls Out in Stages
According to the source report, Google’s stated plan is to offer Carbon first to a subset of cyber defenders, then to trusted partners, developers, businesses and consumers. No dates are given for those steps, and general users should not assume the model is available now.
For organizations considering it, the next useful evidence will be independent or customer-specific testing on their own tasks, alongside clearer availability and pricing terms. Artificial Analysis’ ranking provides a comparison across its test suite; it does not settle how Carbon will perform in a particular company’s workflow.
Key Questions
Is Gemini 4 Carbon the top AI model?
It ranks first on Artificial Analysis’ Intelligence Index v5, with 59 points. That is a result on this index, not proof that Carbon leads every model on every task.
How far ahead is Carbon?
Carbon scores one point above Claude Opus 5.5 and two points above GPT-6 Astra in the reported index results.
Can the public use Gemini 4 Carbon now?
No. The source report says Google is starting with a subset of cyber defenders and plans later stages, but it gives no public release date.
Does Carbon answer more questions correctly than its rivals?
Not on the plain-accuracy comparison reported for AA-Omniscience. Several models score higher on accuracy; Carbon leads that test because its scoring penalizes wrong answers and its reported hallucination rate is lower.
How much does Carbon cost to use?
The report lists an introductory API price of $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20. It also estimates $7.76 per index task at standard prices; those are different measures.
Source: ThorstenMeyerAI.com
