Novel Cognition Analyzes Real-World AI Model Costs
New data reveals total task costs vary significantly between leading artificial intelligence models.
Updated on Oct. 29, 2026 in Artificial Intelligence
Live Poll
Do you trust that AI software providers offer transparent pricing for the tasks you complete?
Novel Cognition has published a comprehensive analysis comparing the true costs of running AI models on standardized tests. The report highlights that despite uniform per-token pricing, total project costs differ due to the varying volume of output text produced by each model.
Why it matters
Reasoning models frequently produce different amounts of output to complete the same task, creating hidden disparities in total billable costs. This analysis helps users predict actual expenditure rather than relying solely on advertised per-million token rates.
Leading models including OpenAI GPT-6.1 Sol and Anthropic Claude Sonnet 5.5 maintain a standard pricing structure of $2 per million input tokens and $10 per million output tokens. Artificial Analysis now utilizes its updated Intelligence Index v5.0 to measure cost efficiency across nine distinct operational test categories.
The players
Novel Cognition
This research firm focuses on the transparency and economic evaluation of large language models.
Artificial Analysis
This entity provides comparative performance and cost benchmarking services for the artificial intelligence industry.
The details
The analysis indicates that Artificial Analysis runs models through a nine-test suite encompassing office productivity, terminal-based coding, and complex scientific queries. Researchers found that while developers often set similar token-based prices, the total cost for a user is driven by how verbosely each model responds to a prompt.
Timeline
26 October 2026: OpenAI released GPT-6.1 Terra.
28 October 2026: Google released Gemini 4 Carbon.
29 October 2026: Novel Cognition performed the pricing analysis.
30 November 2026: Gemini 4 Argon price increase begins.
The Tech Race
This analysis aligns with the ongoing standardization efforts of the Artificial Analysis Intelligence Index v5.0. It marks a critical shift in the sector from simply comparing raw token prices to measuring actual operational efficiency in complex coding and reasoning tasks.
Users can expect significant differences in their monthly cloud compute bills depending on which model they select for automated reasoning tasks. Developers should monitor specific model verbosity to optimize costs when deploying AI agents for high-volume office or coding workflows.
The takeaway
Selecting an AI model requires evaluating total output volume rather than just advertised per-token prices. Users should prioritize models that demonstrate high efficiency on standardized tasks to keep long-term operating costs predictable.
What happens next
Google will implement a price adjustment for Gemini 4 Argon on 1 December 2026, doubling the cost to $4 per million input tokens and $20 per million output tokens.
Further reading
For broader insights into model benchmarking, visit the Artificial Intelligence section.
More information
Review the full AI agent cost analysis report for detailed model performance data.
Source note: This article includes information reported by The Cherry Creek News.
Live Poll
Do you trust that AI software providers offer transparent pricing for the tasks you complete?
