Two of the biggest names in technology made major announcements on the same day. OpenAI introduced GPT-6 Sol and GPT-6 Luna, while Anthropic released Claude Opus 5.5. The launches came within hours of each other and carried a similar message: a powerful model only matters if businesses can afford to run it at scale.

Beyond Benchmark Scores

For a long time, the easiest way to judge a new model was to look at test results. That is changing. Companies now talk about how much useful work a model can finish for every dollar spent, and customers are asking the same question. A model that scores slightly higher but costs far more can lose to a cheaper one that is good enough for the job. Both companies leaned into that idea this time.

OpenAI Adds Two Lower-Priced Options

OpenAI says GPT-6 Sol and Luna draw on technology built for GPT-6 Astra, its most powerful model, but are aimed at faster and cheaper everyday use. Price is the headline change.

Sol costs $2 per million input tokens and $10 per million output tokens through the API. Luna is much cheaper, at $0.10 per million input tokens and $0.50 per million output tokens. OpenAI says these rates are 50 per cent lower than the promotional API pricing for the matching GPT-5.6 models.

The two models have different jobs. Sol is built for complex coding and workflows where the system carries out several steps on its own. Luna is designed for focused, high-volume work where speed and low cost matter most. Both accept text and images, and both have a context window of about 1.05 million tokens, so they can take in very large amounts of material at once.

OpenAI also made a claim about accuracy. In an internal test based on real conversations in which users had flagged errors, the company said GPT-6 Sol made about half as many mistakes as GPT-5.6 Sol. That result comes from OpenAI’s own testing, so it is worth treating with some care until outside groups check it.

Both models are available through OpenAI’s API and are rolling out in ChatGPT Work and Codex for eligible paid users. Luna can also be tried by Free and Go users through the desktop app.

Anthropic Answers With Opus 5.5

Anthropic’s release is Claude Opus 5.5, the first member of its new Claude 5.5 family. The company says it performs at roughly the level of its larger Claude Fable 5.1 model on most tasks, while costing 40 per cent less to run on typical workloads than Opus 5.

The model is priced at $4 per million input tokens and $20 per million output tokens, with cache reads at $0.20 per million tokens. Anthropic adds that it produces output more than 30 per cent faster than Opus 5.

Coding and long-running autonomous work get the most attention. Anthropic says Opus 5.5 beats its earlier models on several coding, computer-use and professional-work evaluations. As always, results published by the developers themselves are not a final verdict on how a model behaves in real jobs.

Safety is the other major theme. Anthropic says the model did better than its predecessors in automated behavioural and alignment testing, and that it holds up better against prompt-injection attacks, where hidden instructions are slipped into content to steer a model off course. Independent safety researchers tested it before release, and it ships with extra safeguards around sensitive cybersecurity and biology capabilities.

More is on the way. Anthropic says Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks, which should fill out the rest of the family at lower price points.

What It Means for Buyers

The timing of these launches points to a change in how the industry competes. Frontier models keep getting more capable, but the bigger contest is now about making that capability cheap and practical enough to use widely. For a company running models around the clock, a lower cost per task can matter as much as a top score on a leaderboard.

Expect customers to compare price sheets closely, test the new models on their own work, and read launch-day claims with a critical eye. Speed, accuracy and reliability on everyday tasks will likely decide who wins business, more than any single chart in a press release.

The coming weeks will be revealing. Sonnet 5.5 and Haiku 5.5 have yet to arrive, and OpenAI’s new pricing will face real-world use as more people try Sol and Luna. If the lower prices hold and the results match the promises, teams that once held back on cost may find far more room to put these tools to work.

I can also put this in a Word or Markdown file if you’d like. I read “AI-related words” as overused, machine-sounding phrasing and left those out, along with underscores and the word “underscore”.