🔍 Read the full analysis: The AI Ranking Names Gemini 4 Carbon #1 By One Point. Read The Fine Print on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Artificial Analysis’ Intelligence Index v5 ranks Google’s Gemini 4 Carbon first with 59 points, one ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The score reflects nine selected tests, not a universal measure of AI ability; Carbon leads four tests, trails rivals on others and is not yet publicly available.
The index combines results from **nine tests** covering knowledge work, coding, science and general capabilities, with each category given **25% of the total weight**. Artificial Analysis says 60% of the v5 evaluation uses private test sets or held-out answers. The tests include agent-style work such as operating in a terminal, building reports and fixing code.
Carbon places first in **four of the nine tests**, according to the source report: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA. It scores 69% on HLE-Diamond, one point above Claude Opus 5.5, and 65% on Terminal-Bench-Science, two points above GPT-6 Astra. Its 78% result on AutomationBench-AA is described as fractionally ahead of its predecessor, Gemini 4 Argon.
The overall ranking is close: **Claude Opus 5.5 scores 58**, GPT-6 Astra 57, GPT-6.1 Sol 55 and Gemini 4 Argon 50. Carbon does not lead every category. The supplied report says rival Claude models outperform it on office-style knowledge work, Claude Opus 5.5 leads on real-world coding, GPT-6 Astra leads on a physics research test, Kimi K3.1 Pro on terminal tasks, and Meta’s Muse Wonder Full on document reasoning.
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
The result matters because benchmark rankings can shape procurement decisions, product comparisons and expectations about which model to try. But the one-point margin makes the **choice of benchmark and task** consequential: a company focused on office documents or coding may care more about Carbon’s losses in those areas than its overall placement.
Price is another part of the comparison. The source report estimates **$7.76 per index task** at standard prices. That is about 23% below Claude Opus 5.5 at $10.04, but around 1.2 times GPT-6 Astra’s $6.24 and about six times GPT-6.1 Sol’s $1.27. These are benchmark-task cost estimates, not a guarantee of what a particular customer will pay; real costs depend on usage and workload.
For now, the ranking has no immediate consumer impact. Google is reportedly giving Carbon first to a **subset of cyber defenders**, then plans a broader sequence of access for trusted partners, developers, businesses and consumers. No dates are provided. The claimed rollout order and pricing come from the supplied report, and readers should distinguish those details from a currently available public product.
Popular right now
TCL QM64L 85-Inch QD-Mini LED 4K Smart Fire TV
Roborock F25 Ultra Steam Wet Dry Vacuum Mop
Vitamix 5200 Professional-Grade Blender, 64 oz. Container, Black
As an affiliate, we earn on qualifying purchases.
How the new index scores models
Artificial Analysis launched **version 5 on October 27**, replacing its earlier leaderboard after rebuilding the index. The v5 design divides its nine tests into four equally weighted blocks: knowledge work, coding, science and general performance. Individual tests have different weights within each block, so the composite score reflects both the test results and the index’s chosen weighting.
The index is best understood as a **comparative score on those nine evaluations**, rather than a universal intelligence rating. Its outcomes can change when the tasks, model settings or weights change. The source report also notes that long, agent-based tasks are costlier to run than simple question-and-answer tests, which affects the price comparisons.
One of Carbon’s reported wins, AA-Omniscience, illustrates why test design matters. Carbon scores 54 on the test, ahead of Claude Opus 5.5 by six points. Yet the report says Carbon’s plain answer accuracy is 58%, below Claude Fable 5.1 at 68%, Claude Opus 5.5 at 66%, and GPT-6.1 Sol and GPT-6 Astra at 62%. The index rewards avoiding unsupported answers: the report gives Carbon an 11% hallucination rate on this evaluation, compared with about 50% for Claude Opus 5.5 and GPT-6 Astra. These figures refer to the test’s scoring context, not every use of the models.
““Number one” on one index is a narrower claim than it sounds.”
— Thorsten Meyer, ThorstenMeyerAI.com
Access, costs and test limits
Carbon’s **public release date is not stated**, and the source does not specify how many cyber defenders or partner organizations will receive early access. It is also unclear when API access will open to developers and businesses, or when consumers might be able to use the model.
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, respectively. The timing and conditions for those price changes are not given. It also does not provide enough detail here to independently assess how the benchmark’s per-task estimates translate to different workloads.
The ranking itself remains limited to the tests and model settings included in v5. The supplied material does not establish how Carbon would perform across every product configuration, deployment or customer task. Its one-point lead over Claude Opus 5.5 is a reported index result; it should not be read as a meaningful advantage in every real-world comparison.
Watch for access and independent testing
The next reported step is Google’s staged access for cyber defenders, followed by trusted partners and then broader groups, but **no timetable has been announced** in the supplied material. Pricing and availability details will need confirmation from Google as those stages proceed.
For organizations considering the model, the relevant next test is performance on their own work: the source report recommends comparing **cost per completed task** as well as quality, rather than relying on the composite rank alone. Further benchmark results or customer evaluations could show whether Carbon’s strengths in science, selected reasoning tests and business workflows outweigh its weaker results in other categories.
For everyday users, the immediate status is straightforward: Carbon is **not publicly available**, so the ranking does not change what they can use today. Its effect on consumer products will depend on when Google expands access and how the model performs in those products.
Key Questions
What is Gemini 4 Carbon’s score?
It scores 59 points on Artificial Analysis’ Intelligence Index v5, according to the supplied report.
How far ahead is Carbon?
Carbon is one point ahead of Claude Opus 5.5, which scores 58, and two ahead of GPT-6 Astra at 57.
Does Carbon lead every test in the index?
No. The report says Carbon places first in four of nine tests and trails competitors in areas including office-style knowledge work, real-world coding, physics research, terminal tasks and document reasoning.
Can the public use Gemini 4 Carbon now?
No. The source report says Google is initially providing access to a subset of cyber defenders, with wider access planned later. It gives no public-release date.
Does the ranking prove Carbon is the best AI model overall?
No. It ranks model performance on nine specified tests with set weights. Results can differ on other tasks, settings or workloads, and the one-point margin is not a guarantee of better performance for an individual user.
Source: ThorstenMeyerAI.com
