Nội dung
- OpenAI launches GeneBench-Pro: measuring AI’s “decision-making” ability, not just correct answers
- Palantir criticizes token-based pricing: a debate over AI’s real value
- Pressure to cut inference costs shows AI is entering a pragmatic optimization phase
- OpenAI and the large net loss problem: AI growth is facing questions about financial efficiency
- A perspective for the Vietnamese market
- References
Recent developments around OpenAI, Palantir, and the cost of inference are highlighting a very notable reality for marketers: AI is not only a race on features, but increasingly a race on operational efficiency and monetization. For brands, agencies, and advertising platforms in Vietnam, this directly affects deployment costs, testing speed, and the “sustainability” of AI campaigns.
-
Key points:
- OpenAI introduced GeneBench-Pro, a new benchmark to test an AI model’s analytical and decision-making capabilities in computational biology research.
- Palantir executives publicly criticized OpenAI and Anthropic’s token-based pricing model, saying the AI industry is heading in the wrong direction on pricing.
- A cited report suggests OpenAI may have found a way to sharply reduce inference costs, reflecting pressure to optimize AI infrastructure.
- Debates over net losses and AI operating costs show the market is entering a phase where it must prove economic efficiency, not just growth speed.
OpenAI launches GeneBench-Pro: measuring AI’s “decision-making” ability, not just correct answers
OpenAI has just announced GeneBench-Pro, a research-level benchmark designed to assess whether a model can handle situations that require real-world judgment in computational biology. What stands out about this benchmark is that it does not stop at testing memorization or the ability to follow preset procedures, but targets more complex decisions: identifying when data reflects true biological patterns, when it is merely noise, and when a result is solid enough to move to the next step. According to OpenAI, GeneBench-Pro expands on GeneBench to cover more difficult tasks in genomics, quantitative biology, and translational medicine. Source: OpenAI.
For marketers, the message here is very clear: AI is being evaluated more and more strictly on reasoning, context, and decision-making, not just the fluency of its output. This reflects the broader trend in the applied AI market — where businesses are no longer satisfied with attractive answers or content, but need systems that can support the right decisions in data-rich, high-risk environments.
Palantir criticizes token-based pricing: a debate over AI’s real value
According to CNBC, Palantir executive Alex Karp attacked the token-based pricing approach of OpenAI and Anthropic, saying that “something has gone very wrong” with this model. While the article does not go deeply into the full technical rebuttal, Karp’s remarks point to a growing wave of skepticism about how the AI industry is making money: charging by the number of tokens processed may be technically easy to understand, but it does not necessarily reflect the value businesses actually receive.
This is a particularly important story for the advertising and marketing industry, which often buys AI tools on a usage basis. As the volume of content, insights, chatbots, data analysis, and media automation grows, token-based pricing can cause costs to rise faster than expected. In other words, the question is no longer “Can AI be used?” but “Can AI be used at what margin?” Source: CNBC.
Pressure to cut inference costs shows AI is entering a pragmatic optimization phase
A source cited by Seeking Alpha says OpenAI has found a way to significantly reduce inference costs, meaning the cost of a model generating output when used in practice. Although the original source is not publicly accessible in this dataset, the very fact that the story about inference costs is drawing attention shows a core issue in the AI market: the more widespread AI becomes, the cheaper it must get to sustain growth.
In advertising, inference is the “heart” of many applications: from ad copy generation and creative support to message personalization, campaign analysis, and internal assistants. If inference costs fall, businesses can expand AI testing across more channels and more layers of workflow. Conversely, if costs remain high, AI is very likely to be limited to showcase use cases rather than becoming a regular operational tool. Reference source: Seeking Alpha.
OpenAI and the large net loss problem: AI growth is facing questions about financial efficiency
CleanTechnica says OpenAI was mentioned in the financial picture with a very large net loss, showing that the cost of developing and operating AI is still creating significant pressure on leading companies. Although the article is commentary and focuses on the economics of the AI wave, the most important message is that market excitement cannot forever obscure questions about capital efficiency and profitability.
For marketers, this is a reminder that adopting AI should not stop at “staying ahead of the trend.” Marketing teams need to measure real impact: how much time is saved, how conversion rates improve, and how content production or media costs are reduced. As AI vendors come under profit pressure, enterprise customers will also face greater pressure to prove clear ROI. Source: CleanTechnica.
A perspective for the Vietnamese market
For the Vietnamese market, this set of news highlights an important shift: AI in advertising and marketing is moving from the “try it and see” stage to the “must prove effectiveness” stage. Vietnamese businesses, especially brands, agencies, and e-commerce platforms, should begin evaluating AI tools against three more practical criteria: usage cost, output quality in Vietnamese-language contexts, and the ability to integrate into existing workflows.
In the short term, changes in benchmarks, pricing, and inference costs may not affect day-to-day campaigns immediately. But in the medium term, they will determine which AI is cheap enough to use at scale, which AI is accurate enough to support decision-making, and which AI remains only at the experimental stage. For Vietnamese marketers, the right priority now is not to chase every new feature, but to build a clear ROI measurement system for each AI application in advertising, content, and customer care.
See more marketing news and guides at https://marketing365.vn.
Follow more updates from Marketing365 to stay up to date with the latest marketing trends.
Read more articles in the Digital Trends category.
This article focuses on AI cost with a perspective for the Vietnamese market.
References
- OpenAI — Introducing GeneBench-Pro
- CNBC — Palantir’s Karp bashes OpenAI, Anthropic token model: ‘Something has gone completely wrong’
- Seeking Alpha — OpenAI finds way to sharply cut inference costs: report (OPENAI:Private)
- CleanTechnica — OpenAI Went From $5.09 Billion Net Loss in 2024 to $38.53 Billion Net Loss in 2025
