🔍 Read the full analysis: The New #1 AI Model Ranking Has Gemini 4 Carbon Ahead 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
Gemini 4 Carbon scored 59 on Artificial Analysis’ Intelligence Index v5, one point ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The result reflects nine specific tests, not a universal measure of AI capability; Carbon is not yet publicly available and is being released first to a limited group of cyber defenders.
Google’s Gemini 4 Carbon now leads the 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 the benchmark results reported October 29. The result puts Carbon at the top of this particular nine-test ranking, but Google has not made the model publicly available: it is starting with a limited group of cyber defenders.
The reported scores are 59 for Gemini 4 Carbon, 58 for Claude Opus 5.5 at maximum reasoning, 57 for GPT-6 Astra at maximum reasoning and 55 for GPT-6.1 Sol at maximum reasoning. Google’s previous model, Gemini 4 Argon, scored 50. The rankings are for specified model settings; they should not be read as a direct comparison of every configuration or product built around those models.
Carbon ranked first on four of the index’s nine tests: HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA. It did not lead the other five. The source report says Claude models performed better on office-style knowledge work, Claude Opus 5.5 led the cited real-world coding comparison, GPT-6 Astra led a physics research test, Kimi K3.1 Pro led a terminal task benchmark, and Meta’s Muse Wonder Full led document reasoning.
The report puts Carbon’s cost at $7.76 per index task at standard prices. That is below Claude Opus 5.5’s reported $10.04, but above GPT-6 Astra’s $6.24 and GPT-6.1 Sol’s $1.27. Google’s introductory API price is listed as $2 per million input tokens and $10 per million output tokens, with later prices of $4 and $20, respectively. These are separate pricing measures from the benchmark’s per-task estimate.
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 Real Trade-Offs
The one-point margin makes the ranking a limited signal, not a clear verdict that Carbon is best for every use. Its position comes from the mix and weighting of nine tests; a company focused on coding, office work, or document analysis may care more about a category where another model leads. The overall score is useful for comparing performance across the index, but it does not replace testing a model on a buyer’s own tasks.
Cost and access also shape the practical impact. Carbon’s reported index-task cost is lower than Opus 5.5’s but higher than the two cited GPT alternatives. And since Google is initially restricting access, the top-ranked score does not mean developers or consumers can switch to Carbon today. For organizations, the result is a reason to evaluate performance and cost per completed task, rather than choosing on rank alone.
The result also highlights how benchmark design affects a model’s apparent strengths. Carbon reportedly does not lead on raw accuracy in the knowledge test, but it scores well on a version that penalizes incorrect answers. That distinction matters for work where a confident wrong response may be more costly than an explicit admission of uncertainty.
Products Worth Considering
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
DJI Osmo Pocket 3 Creator Combo Vlogging Camera with Mic 2, 1 CMOS, 4K/120fps, 3-Axis Stabilization, 128GB Card
Samsung 27 Odyssey OLED G5 (G50SF) QHD QD-OLED Gaming Monitor
As an affiliate, we earn on qualifying purchases.
How Index v5 Produces Its Scores
Artificial Analysis describes its Intelligence Index as a combined score from nine tests grouped into four equally weighted areas: knowledge work, coding, science and general capability. The reported composition gives each area 25% of the total. Within those areas, tests have different weights: for example, AA-Briefcase and FrontierCode 2.0 each account for 15%, while several other tests account for 5% or 10%.
Version 5 launched on October 27, 2026, according to the source report. It says 60% of the index now uses private test sets or held-out answers, intended to reduce the chance that models have encountered test questions during training. Many of the tasks involve longer agent-style work, such as using a terminal, producing reports, or changing code, rather than answering short prompts.
The leaderboard changed quickly: the report says Claude Opus 5.5 was at the top when testers rebuilt the ranking on Tuesday, before Carbon moved ahead the following day. That sequence reflects a revised index and new reported results, not evidence that a model’s capabilities changed overnight.
““‘Number one’ on one index is a narrower claim than it sounds.””
— Thorsten Meyer, source report
Access and Benchmark Limits
Public availability has no announced date in the source material. Google is described as giving Carbon to a subset of cyber defenders first, then trusted partners, followed by developers, businesses and consumers. The timetable, eligibility rules, and whether those stages will happen in that order or on what schedule remain unspecified.
The source material is a report of Artificial Analysis results, but it does not provide a direct statement from Google or the benchmark company confirming every score, price, or rollout detail. It also leaves some test results and comparisons unexplained, including the full performance picture across all nine tests. The index itself cannot establish how Carbon performs on a particular company’s data, workflows, or safety requirements.
Carbon’s reported AA-Omniscience result illustrates another limit: it scores well on a metric that penalizes wrong answers, while the report says several competing models score higher on raw answer accuracy. The best choice may depend on whether a user prioritizes answering more questions correctly, avoiding unsupported answers, speed, or cost. The supplied material says Carbon thinks slowly but does not provide a latency figure or a common measurement method.
Google’s Staged Rollout
Google’s stated rollout plan begins with a small group of cyber defenders, then moves to trusted partners, developers, businesses and consumers. No dates are given for those stages. Further detail on access, API availability, and the change from introductory to later token prices will be needed before customers can judge when Carbon is usable at scale.
For now, the next useful comparison is not another headline position but fuller evidence: results across individual tests, performance on customer workloads, response speed, and actual costs for completed tasks. Buyers considering a change can use the leaderboard to identify Carbon as a candidate, then compare it with available alternatives on their own evaluation set. Everyday users should expect no immediate change while the model remains unavailable to the public.
Key Questions
What is Gemini 4 Carbon’s score?
It scored 59 points on Artificial Analysis’ Intelligence Index v5, according to the source report. Claude Opus 5.5 scored 58 and GPT-6 Astra scored 57 in the cited settings.
Does the ranking show that Carbon is the best AI model for every task?
No. The score combines nine tests, and Carbon reportedly led four of them while other models led the rest. The ranking measures performance on its chosen tests, not every use case.
Can the public use Gemini 4 Carbon now?
Not according to the source report. Google is initially making it available to a subset of cyber defenders, with later stages planned for trusted partners, developers, businesses and consumers. No public release date is provided.
How much does Carbon cost?
The report estimates $7.76 per index task at standard prices. It also lists introductory API pricing of $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20, respectively.
Why did Carbon lead the knowledge test if other models answered more correctly?
The report says Carbon had lower raw accuracy than several competitors on AA-Omniscience but was penalized less for wrong answers. It attributes Carbon’s lead on that test to an reported 11% hallucination rate, compared with about 50% for Claude Opus 5.5 and GPT-6 Astra. Those figures are from the source report and refer to that test’s stated measure.
Source: ThorstenMeyerAI.com
