🔍 Read the full analysis: The One-Point Difference Behind Gemini 4 Carbon’s AI Ranking on ThorstenMeyerAI.com
Get smart everyday buys 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
Artificial Analysis’s Intelligence Index v5 ranks Google’s Gemini 4 Carbon first with 59 points, one point ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The result reflects performance across nine tests, not a universal measure of AI ability; Carbon leads four tests, falls behind on five and is not yet publicly available.
Google’s Gemini 4 Carbon ranks first on Artificial Analysis’s newly updated Intelligence Index v5, scoring 59 points—one ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The ranking places Carbon at the top of this nine-test benchmark, but it does not establish that it is best at every kind of work: the source reports Carbon leads four tests and trails rivals on five, while Google has not yet made the model generally available.
The index gives Claude Opus 5.5 a score of 58, GPT-6 Astra 57, GPT-6.1 Sol 55 and Google’s previous model, Gemini 4 Argon, 50. Artificial Analysis launched version 5 on October 27, according to the source material. The reported Carbon result is for its high-reasoning setting, while the listed competitor scores use the settings specified in the comparison.
Carbon takes first place in four of nine tests: HLE-Diamond, a hard reasoning and knowledge test; Terminal-Bench-Science, which assesses science work in a terminal; AA-Omniscience, which penalizes incorrect answers; and AutomationBench-AA, focused on business workflows. It does not lead the index’s other five tests. Claude models score higher on office-style knowledge work, Claude Opus 5.5 leads on real-world coding, GPT-6 Astra leads on the reported physics research test, Kimi K3.1 Pro leads on terminal tasks, and Meta’s Muse Wonder Full leads on document reasoning.
The price comparison in the source estimates $7.76 per index task for Carbon at standard prices. That is below the cited $10.04 for Claude Opus 5.5 at maximum reasoning, but above GPT-6 Astra’s $6.24 and GPT-6.1 Sol’s $1.27. The source also lists an introductory API rate of $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20 respectively. These are reported prices, not a guarantee of what a particular customer’s workload will cost.
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 Practical Trade-Offs
For companies comparing models, the headline score is a starting point rather than a procurement answer. The index combines varied tasks into one figure, but a business may care most about just one area—such as code quality, office documents, scientific workflows or the reliability of answers. Carbon’s mixed results mean its overall first place does not imply a first-place result on a buyer’s specific task.
Cost also depends on how much work a model completes successfully, not only its listed price or benchmark score. The source’s per-task estimate puts Carbon below Claude Opus 5.5 but above GPT-6 Astra and far above GPT-6.1 Sol. Organizations would need to test their own tasks and account for usage, retries and quality before concluding which option is less expensive in practice.
For individual users, the immediate effect is limited because Carbon is not publicly available. If the source’s description of its strengths holds up in use, the clearest difference may appear on long, difficult tasks rather than routine chat. That remains a forecast, not an independently established user experience.
Products Worth Considering
HP Sprocket Photo Booth Instant Color Printer Bundle
Logitech MX Master 4 for Mac, Ergonomic Bluetooth Mouse with Haptic Feedback, Ultra-Fast Scrolling, USB-C Charging, for MacOS, iPadOS, Windo
Apple iPad Air (4th Generation, 10.9-inch, Wi-Fi, 256GB) – Space Gray
As an affiliate, we earn on qualifying purchases.
How Version 5 Builds Its Score
Artificial Analysis describes the Intelligence Index as a combined score from nine tests grouped into four equally weighted categories: knowledge work, coding, science and general capabilities. Within those categories, individual tests have different weights. For example, AA-Briefcase and FrontierCode 2.0 each account for 15% of the full index, while several other tests carry 5% or 10%.
The source says 60% of version 5 uses private test sets or held-out answers, intended to reduce the chance that models benefit from memorizing publicly available questions. Many tests also involve longer agent-style tasks, such as working in a terminal, preparing reports or modifying code. These design choices shape what the index measures; they do not make it a complete assessment of every use of AI.
The results also depend on model settings. The comparison names Carbon at high reasoning and lists some rivals at maximum reasoning. A different model configuration, test mix or weighting could produce a different order. The score should therefore be read as a result on this index, under these conditions, rather than as a universal ranking of intelligence.
What the Score Cannot Settle
The source is a report from ThorstenMeyerAI.com and does not include a direct statement from Artificial Analysis explaining the result or from Google detailing the rollout schedule. The supplied material reports the scores and test outcomes, but the underlying benchmark data and full methodology are not independently established here. The comparisons should be treated as claims attributed to that report and the benchmark description it provides.
It is also unclear when Carbon will become available beyond the initial cyber-defender group. Google is described as planning access for trusted partners, then developers, businesses and consumers, but the source gives no dates for those stages. The source does not establish how Carbon will perform across customers’ real-world tasks, whether the reported prices will apply to all users, or how its results compare under different model settings.
One benchmark result also needs careful interpretation: Carbon scores highest on AA-Omniscience despite answering fewer questions correctly than several rivals, according to the source. The reported explanation is that the test penalizes wrong answers and Carbon has a lower hallucination rate. That makes the result a measure of cautious answering under that test’s scoring rules, not evidence that Carbon knows more facts overall.
Google’s Staged Access Rollout
Google’s reported plan is to give Carbon first to a subset of cyber defenders, followed by trusted partners, developers, businesses and consumers. No timetable is specified, so the next meaningful update would be a confirmed expansion of access or a public release date.
Until then, companies interested in the model will need to wait for access or rely on the reported benchmark and pricing figures. When testing becomes possible, the most useful comparison will be performance on their own work: answer accuracy, task completion, latency and cost for a finished task. The index ranking alone cannot resolve those questions.
For readers, the near-term development is the benchmark lead, not a change to an available consumer product. Public access, independent real-world evaluations and a clearer rollout timetable remain the developments to watch.
Key Questions
What score did Gemini 4 Carbon receive?
It scored 59 points on Artificial Analysis’s Intelligence Index v5, according to the source. Claude Opus 5.5 scored 58, and GPT-6 Astra scored 57.
Does the ranking show that Carbon is the best model for every task?
No. The reported index combines nine tests, and Carbon leads four while trailing other models on five. The result reflects performance on this benchmark, not every possible use.
Why did Carbon lead the AA-Omniscience test?
The source says the test penalizes incorrect answers. Carbon reportedly had an 11% hallucination rate, compared with about 50% for Claude Opus 5.5 and GPT-6 Astra, allowing it to score well despite lower plain-answer accuracy than several models.
Can the public use Gemini 4 Carbon now?
Not according to the source material. Google is reported to be starting with a subset of cyber defenders and then expanding access, but it has not provided a public release date in the supplied information.
How much does Carbon cost?
The source reports an estimated $7.76 per index task at standard prices and an introductory API rate of $2 per million input tokens and $10 per million output tokens. It says those API rates are later expected to rise to $4 and $20; actual costs depend on usage.
Source: ThorstenMeyerAI.com
