Back to list
Why Token Pricing is a Misleading Metric for AI Value: Insights from Factory’s Eno Reyes
Industry NewsArtificial IntelligenceAI EconomicsFactory

Why Token Pricing is a Misleading Metric for AI Value: Insights from Factory’s Eno Reyes

In a recent discussion, Eno Reyes from Factory challenged the prevailing industry standard of using token prices to evaluate artificial intelligence performance. Reyes argues that judging AI based solely on the cost per token is a misleading approach that fails to capture the actual economic utility of the technology. Instead, he advocates for a shift in perspective toward measuring the 'true cost of finished work.' This shift emphasizes the importance of the final output and the total resources required to achieve a completed task, rather than the raw data units processed. By focusing on the end result, businesses and developers can gain a more accurate understanding of AI efficiency and value, moving beyond superficial pricing models that do not reflect the complexity or quality of the work performed.

Tech in Asia

Key Takeaways

  • Token Pricing is Misleading: Evaluating AI based on the price of individual tokens does not provide an accurate measure of its true value or efficiency.
  • Focus on Finished Work: The industry should prioritize the "true cost of finished work" as the primary metric for judging AI performance.
  • Eno Reyes' Perspective: As a representative of Factory, Reyes highlights that the end result of AI labor is more significant than the granular cost of data processing.
  • Economic Realignment: Shifting metrics from input (tokens) to output (finished work) allows for a better understanding of AI's actual impact on productivity and business costs.

In-Depth Analysis

The Fallacy of Token-Based Evaluation

The current landscape of artificial intelligence is often dominated by discussions surrounding token pricing. Tokens, which represent fragments of words or characters used by large language models to process information, have become the de facto currency for measuring AI usage. However, Eno Reyes of Factory suggests that this focus is fundamentally flawed. When businesses judge AI solely by the price per token, they are looking at an input metric that does not necessarily correlate with the quality or the completion of a task.

A low token price might suggest affordability, but if the AI requires a high volume of tokens to produce a mediocre result, or if it fails to complete the task without extensive human intervention, the perceived savings vanish. Reyes argues that this metric is misleading because it abstracts the AI's performance from its actual utility. In essence, the cost of the raw material (tokens) is being prioritized over the value of the manufactured product (the finished work).

Prioritizing the True Cost of Finished Work

The alternative proposed by Reyes is a transition toward measuring the "true cost of finished work." This approach looks at the total expenditure—including time, computational resources, and necessary iterations—required to reach a final, usable output. Whether the task is generating a software feature, drafting a legal document, or providing a customer service resolution, the value lies in the completion of that specific unit of work.

By focusing on the finished work, stakeholders can evaluate AI based on its ability to deliver results. This metric accounts for the efficiency of the model in a way that token pricing cannot. For instance, a more expensive model (in terms of token price) might actually be more cost-effective if it can complete a complex task in a single pass with minimal tokens, whereas a "cheaper" model might require multiple prompts and a higher total token count to reach the same goal. Reyes emphasizes that the end-to-end cost of achieving a result is the only figure that truly matters in a professional or industrial context.

Industry Impact

The shift from token-centric pricing to a work-centric evaluation model has significant implications for the AI industry. For AI providers, it encourages the development of models that are not just cheaper to run on a per-unit basis, but are more effective at solving problems autonomously. This could lead to a competitive environment where "completion rates" and "accuracy per task" become more important than "tokens per dollar."

For enterprises, this perspective simplifies the ROI calculation for AI integration. Instead of navigating complex pricing tiers based on technical units they may not fully understand, business leaders can assess AI tools based on the cost of replacing or augmenting specific human workflows. This aligns AI procurement more closely with traditional business logic, where the value of a service is determined by the quality and reliability of the output. Ultimately, Reyes' insights suggest that as the AI market matures, the metrics used to define success must evolve from technical inputs to tangible, finished outcomes.

Frequently Asked Questions

Question: Why is token pricing considered a misleading way to judge AI?

Token pricing is considered misleading because it focuses on the cost of raw data processing rather than the value of the final result. A low price per token does not guarantee that the AI will be efficient or effective at completing a specific task, often ignoring the total volume of tokens or human oversight required to get a usable output.

Question: What does "true cost of finished work" mean in the context of AI?

The "true cost of finished work" refers to the total expense and resources required to take a task from start to completion. This includes the cumulative cost of all tokens used, the time taken, and any other operational costs involved in producing a final, high-quality result that meets the user's needs.

Question: How does focusing on finished work change how businesses choose AI tools?

When businesses focus on the cost of finished work, they prioritize the AI's ability to deliver a final product efficiently. This leads them to choose tools based on their success rate and total cost-to-completion rather than just looking for the lowest price per token, ensuring better alignment with actual business goals and productivity.

Related News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini
Industry News

How Google Duplex Evolved to Call Businesses on Your Behalf in Gemini

A Google Research Blog entry titled “How Google Duplex evolved to also call businesses for you in Gemini” describes its subject as the evolution of Google Duplex to support calls to businesses on a user’s behalf in Gemini. The entry is dated November 20, 2026, and the supplied article text consists only of the label “Generative AI.” The headline identifies the capability and connects it to Duplex, but the available text does not explain how the system evolved, how users request a call, or where the feature is available. The headline establishes the topic, but the supplied material does not provide a technical account or rollout details.

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop
Industry News

How AI and Lean Are Helping Automate Formal Verification to Fight Software Slop

In Latent Space, Richard MacManus examines formal verification as an emerging way to fight what the headline calls “slop.” The central claim is that proving software correct with mathematics has traditionally required specialist work, and that AI and Lean are now helping automate that process. The distinction is important: the reported change concerns how the work of proving correctness is done, not what counts as a proof. The supplied excerpt presents automation as assistance, without saying that it replaces specialists or completes proofs independently. It also does not specify which parts of the process AI and Lean help automate, how much work they save, or which software has been verified this way.

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis
Industry News

AI and Human Claim Solutions to the 167-Year-Old Riemann Hypothesis

TechCrunch AI reports that AI claimed a solution to the Riemann hypothesis, a mathematical problem the headline describes as 167 years old. A human also put forward a competing solution. The report, written by Tim Fernholz and published on November 20, 2026, describes mathematicians as astonished by the AI-driven result. The supplied excerpt does not explain either proposed solution, identify the AI system or the human claimant, or say whether either claim has been verified. The distinction between a claimed solution and a confirmed one is central to reading the report: the excerpt establishes that two claims are being reported, but it does not establish that either has resolved the hypothesis.