🔍 Read the full analysis: A Close AI Ranking Puts Gemini 4 Carbon At #1 By One Point on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Artificial Analysis’ Intelligence Index v5 places Google’s Gemini 4 Carbon first with 59 points, one ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The narrow lead reflects results across nine tests, not a universal measure of model ability; Carbon is not yet publicly available, and its cost and uneven benchmark results may matter more than its rank for prospective users.
Google’s Gemini 4 Carbon ranks first on Artificial Analysis’ Intelligence Index v5 with 59 points, one point ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra, according to figures published by ThorstenMeyerAI.com on October 29. The result puts Carbon at the top of this particular nine-test benchmark, but the model is not yet publicly available and its narrow lead does not mean it performs best on every task.
The reported scores place Claude Opus 5.5 at 58, GPT-6 Astra at 57, GPT-6.1 Sol at 55 and Google’s earlier Gemini 4 Argon at 50. Carbon is listed at a high-reasoning setting, while the competing Claude and GPT models are listed at their maximum settings. The ranking therefore compares results under the stated configurations, rather than establishing a single performance result for every possible setting.
Carbon reportedly ranks first on four of the index’s nine tests: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA. It does not lead the other five. Claude models score higher on office-style knowledge work, Claude Opus 5.5 leads the cited real-world coding test, GPT-6 Astra leads on a physics-research test, and Meta’s Muse Wonder Full leads on document reasoning. Kimi K3.1 Pro is reported to lead on another terminal-task benchmark.
The source puts Carbon’s cost at $7.76 per index task at standard prices. That is below the cited $10.04 for Claude Opus 5.5 at maximum settings, but above GPT-6 Astra’s $6.24 and well above GPT-6.1 Sol’s $1.27. Google’s reported introductory API rates are $2 per million input tokens and $10 per million output tokens, with later rates of $4 and $20, respectively. The report does not give a public launch date.
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 Narrow Lead, With Trade-Offs
The ranking may help businesses identify models to evaluate, but the score alone is not a purchasing recommendation. Carbon’s one-point margin is small, and its test-by-test results differ: a company that relies on office documents or coding may care more about those specific comparisons than the overall index position. The index measures performance on its selected tasks, not every workflow a customer might run.
Cost also changes the practical comparison. The reported $7.76 per task is lower than the Claude Opus 5.5 figure in the same comparison, but higher than the two cited GPT alternatives. Since the index includes long agent-style work, the price of completing a task may be more useful to buyers than a model’s token rates or rank alone. The source recommends testing models on a company’s own work before switching; the data provided do not establish how Carbon will perform across individual customers’ tasks.
For ordinary users, the immediate effect is limited: Carbon cannot yet be accessed publicly. If the reported rollout proceeds, its strongest results appear most relevant to demanding reasoning, science and business workflows. The source suggests everyday chat may show less difference, but that is an interpretation, not a confirmed assessment of public user experience.
Popular right now
De'Longhi Magnifica Evo Fully Automatic Espresso Machine with Manual Milk Frother
Bose QuietComfort Wireless Bluetooth Headphones with Active Noise Cancellation, Mic, USB‑C Charging, Deep Bass, 24‑Hour Battery – White Smok
KitchenAid Artisan 5-Quart Tilt-Head Stand Mixer, Almond Cream
As an affiliate, we earn on qualifying purchases.
How Index V5 Scores Models
Artificial Analysis launched Intelligence Index v5 on October 27, according to the source. It describes the index as a combined score from nine tests grouped into four equally weighted areas: knowledge work, coding, science and general performance. The tests include tasks such as building reports, using software tools, working in a terminal and answering difficult knowledge questions. The source says 60% of the version 5 assessment uses private test sets or held-out answers.
That design makes the index a snapshot of performance on a defined set of tasks and settings, rather than a general verdict on which model is “the smartest.” The source says Claude Opus 5.5 led the leaderboard on Tuesday, before Carbon took the top position a day later. That sequence reflects the reported ranking at those times; it does not establish how often the board changes or whether the models have been tested under identical conditions beyond the stated configurations.
One reported result illustrates why the overall score needs detail. Carbon scores 58% on plain accuracy in AA-Omniscience, below several listed Claude and GPT models. Yet it leads that test’s scoring measure, which penalizes wrong answers, with 54 points. The source reports an 11% hallucination rate for Carbon, compared with about 50% for Claude Opus 5.5 and GPT-6 Astra. In this benchmark, fewer incorrect guesses can raise a score even when another model answers more questions correctly.
“Think of it as a decathlon score for AI models.”
— ThorstenMeyerAI.com, describing the Intelligence Index
Release Timing and Benchmark Limits
Several details remain unconfirmed in the supplied report. Google has not given a date for public availability; the reported sequence is an initial rollout to some cyber defenders, followed by trusted partners, then developers, businesses and consumers. The source does not specify how large those groups will be, how long each stage will last, or whether access and pricing will differ by customer.
The ranking figures and performance claims are attributed here to the source report, which cites Artificial Analysis results. The supplied material does not include direct statements from Artificial Analysis about its methodology or from Google about Carbon’s release. It also does not provide uncertainty ranges, repeat-run results or enough detail to independently compare the models’ test settings and operating costs. Carbon’s lead should therefore be read as a reported result on this index, not proof that it is the best choice for every use.
Google’s Staged Rollout
According to the report, Google plans to provide Carbon first to a subset of cyber defenders, then to trusted partners, developers, businesses and consumers. No dates are given for those stages. Until access expands, customers cannot test the model directly through the reported public channels or assess whether its benchmark advantages carry over to their own work.
Further useful developments would include Google’s release schedule, confirmation of API availability and prices, and independent results from users applying Carbon to real tasks. For buyers comparing models, the next step is to evaluate the relevant tasks and total cost of completed work once access is available, rather than relying on the overall leaderboard score by itself.
Key Questions
What is Gemini 4 Carbon’s score?
It is reported to score 59 points on Artificial Analysis’ Intelligence Index v5, one point ahead of Claude Opus 5.5 at 58.
Does Carbon lead every test in the index?
No. The source says Carbon ranks first on four of nine tests and trails other models on office-style knowledge work, real-world coding, physics research, document reasoning and a terminal-task test.
Can the public use Gemini 4 Carbon now?
Not according to the October 29 report. It says Google is starting with a subset of cyber defenders and has not announced a date for general availability.
Is Carbon cheaper than competing models?
It depends on the comparison. The reported $7.76 per index task is below Claude Opus 5.5’s $10.04, but above GPT-6 Astra’s $6.24 and GPT-6.1 Sol’s $1.27.
Does first place mean Carbon is the best model for every user?
No. The index combines results from nine selected tests. Carbon’s overall lead is narrow, and other models perform better on several individual tasks, so suitability depends on the work and its cost.
Source: ThorstenMeyerAI.com
