TL;DR
Artificial Analysis data accessed November 17, 2026, gives Claude Fable 5.2 the highest Intelligence Index score among the compared models, but Gemini 4 Carbon leads on several workplace tests and costs less per task. The results vary by task, and Fable’s higher answer accuracy comes with a higher measured hallucination rate when it does not know an answer.
Independent benchmark results accessed November 17, 2026 put Anthropic’s Claude Fable 5.2 ahead of Google’s Gemini 4 models on an overall intelligence index, while Gemini 4 Carbon leads several work-specific tests and costs less per benchmark task. The comparison offers businesses evidence for choosing by workload, but does not identify one model as best for every job.
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
On the Artificial Analysis Intelligence Index, Fable 5.2 scored 62 at High effort and 65 at Max. Gemini 4 Carbon scored 60 at High, Hydrogen scored 54 at its highest setting, and Argon scored 50 at High. At the matched High setting, Artificial Analysis estimated the cost of one index task at $9.08 for Fable 5.2 and $5.48 for Carbon. Its Max setting cost $26.54 per task.
The task-level results split. Fable 5.2 at High scored 49.1% on repository coding, against Carbon’s 40.6%, and 83.3% on long-context work, against 79.3%. Carbon scored higher on terminal work, professional documents and business software automation: 65.2%, 28.4% and 77.8%, respectively. Fable scored 59.6%, 21.8% and 71.4% on those tests.
Artificial Analysis reported Fable’s answer accuracy at 77.2%, compared with 58.3% for Carbon. Its measured hallucination rate was also higher: 58.0% versus 11.4%. The figures describe performance in the cited evaluations; they do not establish how either model will behave on every company’s data or workflow.

Match Model Strengths to Work
For organizations selecting an AI model, the results put a price on different strengths. Teams focused on coding in real repositories or long documents may find Fable’s benchmark lead relevant. Teams using terminal tasks, document preparation or business software automation may give more weight to Carbon’s higher scores in those categories and its lower measured task cost.
The reliability figures raise a separate deployment question. Fable answered more test questions correctly overall, while Carbon was less likely to give a wrong answer in cases measured as ones it did not know. Businesses should consider both outcomes when deciding where model responses need review, especially for tasks where an incorrect confident answer could cause problems.
These are benchmark comparisons, not guarantees of workplace productivity or savings. A company’s own prompts, tools, data controls and review process can change both quality and cost. The results support testing models on representative tasks before committing a broad workload to one system.
Products Worth Considering

Logitech MX Mechanical Wireless Illuminated Keyboard, Tactile Quiet, Graphite

Anker SOLIX C1000 Gen 2 Portable Power Station with 200W Solar Panel

Samsung 27-Inch Odyssey OLED G5 QHD Gaming Monitor (G50SF)
As an affiliate, we earn on qualifying purchases.
Four Days of Gemini and Claude Releases
Anthropic released Claude Fable 5.2 on November 9, describing it as a model for ambitious, open-ended technical work. Google opened Gemini 4 Carbon to paying users on November 10, then released Gemini 4 Hydrogen to all users on November 12. Google also began using Hydrogen in Google Search, according to the source material.
Gemini 4 is a model family rather than a single offering. Google describes Carbon as its most capable reasoning model and recommends the more efficient Argon for most work. Hydrogen is positioned for rapid execution across agent workflows. The compared models have context windows of up to one million tokens, according to the cited model information, though their input formats differ.
Prices also vary by model and date. Artificial Analysis’s task-cost figures put Carbon at $5.48 per index task and Argon at $4.97, based on Google’s introductory prices. Google’s listed prices for Carbon and Argon double on December 1, 2026, from $2 to $4 per million input tokens and from $10 to $20 per million output tokens. Hydrogen is listed at $1.50 per million input tokens and $7.50 per million output tokens. Fable 5.2 is listed at $8 and $40, respectively. Token prices alone do not show total task costs because models can use different numbers of tokens while reasoning.
““the most ambitious, open-ended technical work””
— Anthropic
Benchmark Results Leave Gaps
The figures come from named Artificial Analysis evaluations, not a controlled study of business outcomes across companies. The source material does not provide the full test protocols or enough detail to assess how closely each benchmark matches a particular organization’s work. Results on an individual company’s tasks may differ.
Cost per task depends on the evaluation and model settings, while actual bills depend on usage, token volume and pricing. Carbon and Argon’s stated introductory prices change on December 1, so cost comparisons based on the earlier rates will not describe later use. The source also does not establish whether the measured accuracy and hallucination rates predict error rates in specific business applications.
Test Workloads Before Switching
Businesses weighing the models can compare them on a defined set of their own coding, document, terminal and software tasks, then track answer quality, review needs and cost under the settings they plan to use. Google’s Carbon and Argon price change is scheduled for December 1, 2026, making the date relevant to teams estimating ongoing costs. No further independent results or company-specific evaluations are included in the source material.
Key Questions
Which model scored highest overall?
Claude Fable 5.2 scored highest among the listed models on the Artificial Analysis Intelligence Index, with 62 at High effort and 65 at Max. Gemini 4 Carbon scored 60 at High.
Is Gemini 4 Carbon better for work?
It scored higher than Fable 5.2 on the cited terminal, professional-document and business-software tests. Fable scored higher on repository coding and long-context work, so the results depend on the task.
Which model costs less per benchmark task?
In the cited index evaluation, Gemini 4 Carbon cost $5.48 per task at High effort, compared with $9.08 for Fable 5.2 at High. The Carbon estimate uses Google’s introductory price, which is scheduled to double on December 1, 2026.
What do the reliability results show?
Artificial Analysis measured higher answer accuracy for Fable 5.2 than Carbon, at 77.2% versus 58.3%. Its reported hallucination rate was also higher, at 58.0% versus 11.4%. These benchmark measurements do not guarantee performance on a company’s own tasks.
Source: Thorsten Meyer AI
