🔍 Read the full analysis: The Gemini 4 Carbon AI Ranking Puts It At #1—With A One-Point Margin on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Artificial Analysis’s 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 ranking is based on nine tests, and Carbon leads four while losing five. Google says it is initially making the model available to a limited group of cyber defenders; a public release date has not been given.
Google’s Gemini 4 Carbon now ranks first on the Artificial Analysis Intelligence Index v5, with 59 points—one ahead of Claude Opus 5.5 and two ahead of GPT-6 Astra. The ranking, reported October 29, followed the benchmark’s rebuild a day earlier; it is a narrow result across nine tests, not a general finding that Carbon is best at every task. Google says access is currently limited to a subset of cyber defenders, with no public release date specified.
The index combines results from nine tests grouped into knowledge work, coding, science and general capabilities, each assigned 25% of the overall score. Carbon takes first place in four individual tests: hard reasoning and knowledge questions, science workflows run in a terminal, a knowledge test that penalizes wrong answers, and business workflows across software tools. It finishes behind other models on the remaining five.
In the source’s score table, Claude Opus 5.5 scores 58, GPT-6 Astra 57, GPT-6.1 Sol 55 and Google’s earlier Gemini 4 Argon 50. Carbon’s performance is mixed across tasks: it scores 69% on HLE-Diamond, one point above Opus 5.5, and 65% on Terminal-Bench-Science, two points above GPT-6 Astra. On office-style work, Claude Sonnet 5.5 and Opus 5.5 score higher; Opus also leads Carbon in real-world coding, while other models lead on terminal tasks and document reasoning.
The source reports an estimated $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. Google’s introductory API price is listed as $2 per million input tokens and $10 per million output tokens, later rising to $4 and $20 respectively. Those prices and the index-task estimate are different measures and should not be treated as directly interchangeable.
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 result gives Google a strong showing on a benchmark that combines several demanding tasks, but a one-point margin offers little basis for assuming Carbon will outperform competitors across a company’s full workload. The index itself shows meaningful differences: Carbon leads on some reasoning, science and business-workflow tests, while Claude models do better on office-style work and coding, and other systems lead in additional categories.
For businesses, the practical issue is performance on their own tasks at an acceptable cost—not just an overall leaderboard position. The reported per-task estimate also varies substantially among the leading models. A team deciding whether to adopt Carbon would need to compare output quality, reliability, speed and total cost on its own workflows. Those considerations matter particularly for long-running agent tasks, which the index is designed to test and which can consume more resources than a short chat request.
For individual users, the ranking does not mean an immediate product change. Carbon is not publicly available according to the source material, and Google has not supplied a date for broader access. The benchmark can inform expectations about selected difficult tasks, but it does not establish that ordinary users will see a difference in everyday chat once the model is released.
Products Worth Considering

Bose SoundTouch 30 Wireless Speaker, Black

DJI Mini 5 Pro Fly More Combo with DJI RC 2, 1-Inch CMOS

Bose QuietComfort True Wireless Earbuds with Active Noise Cancellation, Bluetooth 5.3, USB-C Charging, and IPX4 Water Resistance
As an affiliate, we earn on qualifying purchases.
A Rebuilt Nine-Test Benchmark
Artificial Analysis launched Intelligence Index v5 on October 27, according to the source. Its revised design assigns equal weight to four categories: knowledge work, coding, science and general tasks. The tests include writing and software workflows, code and terminal tasks, science research processes, and questions that reward careful handling of uncertainty. This mix helps explain why the overall score can differ from a model’s performance on any single familiar benchmark.
The source says 60% of the index uses private test sets or held-out answers, intended to reduce the chance that models have simply memorized questions. Many assessments require models to act as agents—for example, working in a terminal, producing reports or changing code—rather than only answering isolated prompts. That makes the results more focused on extended tasks, while also making runs more resource-intensive.
The sequence of rankings is brief: Claude Opus 5.5 initially topped the rebuilt leaderboard on Tuesday, then Carbon held the lead a day later at 59 points. The index is a score for specified models at specified settings on its chosen tests. Its results do not cover every use case or establish a universal ranking of AI capability.
Access, Costs and Benchmark Limits
There is no announced public-release date in the source material. Google’s stated rollout sequence is limited cyber defenders first, followed by trusted partners, developers, businesses and consumers, but timing for those stages is unspecified. The source does not provide a way to verify the size of the initial group or the conditions for moving between stages.
The supplied material also does not fully document how the index scores are combined, the number of runs behind each result, uncertainty ranges, or whether the one-point gap is statistically meaningful. The ranking is reported as a point score, but that small difference should not be read as a proven performance separation across all applications. The source excerpt’s description of the coding-agent comparison is cut off, and its quoted statement cannot be completed from the available text.
Finally, price comparisons depend on settings and usage. The stated task-cost estimate and token prices do not show what a particular customer would pay for a finished job, nor do the supplied details quantify Carbon’s latency. The source characterizes it as slow, but gives no timing measurement or comparison methodology.
Google’s Staged Rollout
Google’s stated next step is to provide Carbon to a subset of cyber defenders, then expand access to trusted partners, developers, businesses and consumers. No dates are given for those phases. Until wider access begins, the index score is the main development reported here, rather than a change users can immediately test in Google products.
Artificial Analysis may publish more detail about the benchmark and its model results, while Google’s future rollout announcements could clarify availability and pricing. For organizations evaluating the result, the next useful evidence would be testing Carbon on their own workloads and comparing the cost and quality of completed tasks with alternatives. The source provides no confirmed schedule for either further benchmark updates or public access.
Key Questions
What does Gemini 4 Carbon’s #1 ranking mean?
It means Carbon scored 59 points on Artificial Analysis’s Intelligence Index v5, one point ahead of Claude Opus 5.5. The score combines nine tests and does not establish that Carbon is the strongest model for every task.
Which tests does Carbon lead?
The source reports first place for Carbon on HLE-Diamond, Terminal-Bench-Science, AA-Omniscience and AutomationBench-AA. It loses to other models on five of the nine tests.
Can the public use Gemini 4 Carbon now?
Not according to the source material. Google is starting with a subset of cyber defenders, then plans to expand access in stages. No public launch date is listed.
Is Carbon cheaper than competing models?
It depends on the comparison. The reported estimate is $7.76 per index task, below Claude Opus 5.5’s $10.04 but above GPT-6 Astra’s $6.24 and GPT-6.1 Sol’s $1.27. Actual costs depend on usage and pricing settings.
Does Carbon answer more knowledge questions correctly than its rivals?
No. The source says several Claude and GPT models have higher raw accuracy on the knowledge questions. Carbon wins AA-Omniscience because that test also penalizes wrong answers, and Carbon is reported to have a lower hallucination rate in that comparison.
Source: ThorstenMeyerAI.com
