MARKETS

Why OpenAI is resetting frontier AI prices

OpenAI is usually the AI industry's trendsetter. Its latest focus is all about making tokens more affordable. 

Last week, the company announced that it would cut prices of GPT-5.6 Sol, its highest performing model, by more than 20% for developers for the next three months. The model is now priced at $4 per million ​input tokens and $20 per million output tokens for ‌standard ⁠short-context use, compared to previous pricing of $5 per 1 million ​input tokens and $30 per 1 million output tokens. 

It's also far below the costs of Fable 5, the most powerful model from rival Anthropic, which runs at $10 ​per ⁠million input tokens and $50 per million output tokens. Sol is now also cheaper than Opus 5, Anthropic's most recent model, which comes out to $5 per ​million input tokens and $25 per million output tokens.

It's not the first time OpenAI has slashed prices for its models. Last month, the company chopped prices for GPT-5.6 Terra, its mid-tier model, and GPT-5.6 Luna, its lightweight, efficient model, by 20% and 80%, respectively. 

Though these price cuts help it stay competitive with Anthropic, they also may be edging the company beyond cheaper Chinese alternatives: In a post on X, Tomasz Tunguz, venture capitalist at VC firm Theory, noted that OpenAI's Luna is price competitive with DeepSeek's latest model intelligence-to-dollar. 

Additionally, according to benchmark platform Arena, Sol's price decrease has improved its performance in agentic benchmarks for coding and work.  

These price cuts come as enterprises begin to seriously question their AI spending, with agents poised to increase inference costs up to fivefold, according to recent Gartner research. And the evidence is already starting to pile up that companies are shifting away from implementing frontier AI at any cost: 

• In mid-August, data from payments platform Ramp indicated that most of business spending on Anthropic models wasn't on its ultrapowerful model Fable 5, comprising only 6% of token spend. 

• For comparison, OpenAI's GPT-5.6 Sol comprised 25% of tokens used by OpenAI Ramp customers. 

• Meanwhile, Anthropic's less powerful Opus 5 has surpassed Fable 5 in business spending.

Though OpenAI tends to be one of the industry's tastemakers, it may also be reading the tea leaves from the fact that enterprises are pulling back spending on powerful frontier models, and are largely comfortable settling for efficiency over capability for many tasks. It's why small models, domain-specific models, and open-source alternatives are gaining so much traction. However, as getting enterprises to adopt frontier intelligence is a major part of the company's strategy, these discounts might be a way to get people to see beyond the price tag. Even if these companies don't need Thor's hammer to kill a fly, by selling Thor's hammer on the cheap, it seems that OpenAI wants them to give it a shot anyway. Plus, with OpenAI putting its best models on sale, perhaps Anthropic will follow in its footsteps. 

TOGETHER WITH CELONIS

Don't bet your business on a fortune cookie

Eliminate operational blind spots. Give AI the context it needs to succeed.

AI models are brilliant, but they’re probabilistic. Without a ground truth of your operational reality, trusting AI is like trusting a fortune cookie.

The Celonis Context Model provides this operational reality through a dynamic, real-time, system-agnostic digital twin of your operations. It translates your business into a language that AI understands.

Give your AI the context it needs to drive meaningful ROI. Meet the Context Model.

RESEARCH

What America’s most AI-exposed cities reveal

Across the country, workers are watching AI-powered automation creep into their industries and wondering if and when their jobs could be in danger. But where you live may determine how much you have to fear.

Indeed developed a metric that measures how exposed each metro's job composition is to AI by combining the jobs each metro advertises with the AI exposure of those skills, using data from both the Indeed Hiring Lab's 2025 AI at Work Report and its GenAI Skill Transformation Index. The findings, unsurprisingly, showed that metros built around tech and knowledge work are the most vulnerable. 

Across US metro areas, the average exposure score was 44, with a range of 40 to 60. The five most-exposed metros are driven by software and data work: 

• San Jose, CA: 59 

• Seattle, WA: 57

• Washington, DC: 54

• San Francisco, CA: 53 

• Austin, TX: 52

On the other hand, metros whose economies lean into hands-on work were less exposed. This supports the same conclusion drawn by many job reports and experts alike: labor that relies on physical work is the safest type of job in the AI era. 

As we covered last week at The Deep View, this is already influencing the next generation of the labor force, with Pew Research Center finding that, for the first time, a majority of adults under 30 (55%) say they’re more concerned than excited about AI, and other research showing that Gen Z is more open to careers in trades.

Because the Indeed report findings were based on data examining how AI could change the way certain skills are performed, it is important to note that the jobs most exposed don't necessarily equate to the jobs that will be replaced by AI, but rather to those with the highest potential to be transformed by AI. How that exposure translates into job loss depends entirely on how the employer chooses to adopt the technology. 

With that in mind, An Nguyen, an economist at Indeed Hiring Lab, has advice: 

"For job seekers, it can be a starting point for which skills to build. A high exposure score means more of the work could be reshaped, or augmented, by AI, not that the job will go away," Nguyen told The Deep View. "In other words, exposure is two-sided: it flags where the work is likely to change and where adapting early pays off."

Keeping in mind the metric this report measures, which is skill exposure to the possibility of AI transformation, the key point here is that it is just looking at how well AI can do individual tasks, for example, coding or writing. This report doesn't measure the human element in how those tasks are applied in the workplace, such as how they are synthesized through real-world experiences that cannot be easily replicated by a chatbot, including leadership, critical thinking, and reasoning. If anything, to me, a lot of this data is also reinforcing one thing: the importance of developing human skills that AI cannot replicate.

TOGETHER WITH CRUSOE

$5 in free credits, on us.

Every Crusoe account now comes with $5 in free credits for Crusoe Intelligence Foundry. Use them for Serverless Inference or Serverless Fine-Tuning, whether there's a new open model you've been wanting to try or a workload you're ready to run.

No new commitment, no setup required. Credits apply automatically to your account and are ready whenever you are.

Claim your $5 in credits, start building with Crusoe.

ENTERPRISE

Why companies want to own their intelligence

There's lots of talk about companies "owning their own intelligence." Thomson Reuters is one of the first enterprises to pull it off.

On Monday, Thomson Reuters, the content and technology company, launched Thomson, its first proprietary large language model. Thomson was developed in-house, using an open-source model as its foundation and then trained on decades of proprietary content, including expertise from Westlaw, Practical Law, Checkpoint, and Reuters. The model is specifically designed for professional work and emphasizes the importance of domain expertise. 

Because of the way the model was built, Thomson Reuters fully controls it, which gives the company a better understanding of the intelligence embedded in the model, reduces dependence on another company's AI roadmaps, and, most importantly, reduces costs. The company reports that the endeavor cost $40 million in talent and compute, which is comparatively low compared to what frontier labs spend on training. 

"A real opportunity is for a company like Thomson Reuters to take those frontier open-source models and use them as starting points for our training and development with our unique data and compute. And in that way, we're able to train highly capable AI systems at a fraction of the cost that it would be to train one from scratch," said Joel Hron, Thomson Reuters CTO, in a briefing with the media. 

So far, it has been trained on only 10% of Thomson Reuters' content, but the company says the goal isn't just to keep feeding it more content, but to turn more of it into targeted training data that improves the model. 

The company reports that its early evaluations put Thomson on par with the latest frontier models across a range of tasks. It also attributes the quality of its content as the differentiator from proprietary models. When comparing how GPT-5.4, Sonnet 5, and Thomson 1.0 Large performed with web content only versus Thomson Reuters content only, there was, in every instance, some improvement in factuality or completeness. It also said that in its early evaluations, Thomson performed on par with the latest frontier models across a range of tasks.

At launch, it will be available in CoCounsel Legal, Thomson Reuters' legal AI platform built on its Fiduciary-Grade AI standard, which is designed to meet the requirements of highly regulated industries such as law, tax, and accounting.

With the industry's focus on improving efficiency, there has been a strong emphasis on finding alternative ways to reduce costs when deploying models without compromising results. A key way to do so is to use domain-specific models, which are typically smaller and therefore require fewer tokens. They also tend to be faster and more accurate because they are working from a smaller, more specific data set. These domain-specific models are also key players in model routing, which promises to save users money by tapping into them when useful and is one of the industry's hottest trends. In Thomson Reuters' case, beyond saving money, a major advantage is being able to use its trove of data to inform the AI. However, the interesting thing is that the company has previously entered into licensing deals with AI labs such as Meta and Microsoft. We have to wonder if the company will take a different approach to content licensing now that it has its own proprietary model.

LINKS

• Tech firms shaped US school supply chains via funding, reach

• Texas Gov. Abbott says data centers "dug their own grave" over opposition

• New York Times reportedly tests AI-generated search summaries

• Nvidia to spend $6 billion on bolstering US open source ecosystem

• Business schools prep students for growing role: Chief AI Officer

• Nvidia reportedly in talks to invest in Perplexity at $30 billion valuation

• Codex: OpenAI gives users a full reset of the usage for all paid subscriptions.

• Wan 3.0: Alibaba launches its latest AI model

• ElevenLabs CLI v1: developers and coding agents can now manage ElevenLabs workflows directly in the terminal

• Runway Ruby: New model that converts SDR video up to 16-bit HDR in ProRes and EXR sequences

• Claude: Enterprise-managed auth for MCP connectors is now generally available

• Nvidia: Solutions Architect, AI Models

• Hilbert: AI Engineer - Core

• Kirkland & Ellis: Innovation AI Developer

• Code Red Partners: Artificial Intelligence Engineer

GAMES

POLL RESULTS

Do you think that frontier models will eventually become commoditized?

Yes (48%)
Somewhat (29%)
No (12%)
Other (11%)