Anthropic's price signal is blunt: Claude Opus 5 is cheaper because enterprise buyers are watching every token, while Fable 5 is capped because the best model still isn't cheap to serve.
The Financial Times reported that Anthropic's most advanced model, Claude Fable 5, has struggled to win the spending share a frontier model usually gets after launch. Only 11% of U.S. customer spending went to Fable 5, according to the FT, as buyers leaned on cheaper tools from Microsoft, Amazon, Google and newer Chinese rivals instead of paying the highest Claude rate. That is the story. If you're buying AI for a company, the smartest model in the menu is no longer the automatic choice.
Anthropic's answer arrived on July 24, 2026. Axios reported that Claude Opus 5 launched at $5 per million input tokens and $25 per million output tokens, with Anthropic pitching it as close to Fable 5 on many tasks at half the cost. Anthropic's own pricing page lists Fable 5 at $10 per million input tokens and $50 per million output tokens. The gap is easy to understand when finance teams start asking what a coding assistant or a legal review workflow actually costs at scale - never mind an internal research tool.
Look back one generation and the reset is even clearer. Anthropic's documentation listed Claude Opus 4.1 at $15 per million input tokens and $75 per million output tokens. Opus 5 now sits at one third of that input and output price. That isn't a premium line holding firm. It's a company admitting that enterprise demand has a ceiling.
The cheaper model now has the sharper pitch #
Opus 5 also ships with a control that tells you exactly who Anthropic is trying to win back: effort. In Anthropic's developer documentation, the effort setting lets customers choose low, medium, high, extra high or max, changing how many tokens Claude spends before it answers. Low is for simple, high-volume work. Medium is for saving cost without giving up too much quality. High is the default. Max is for the tasks where you want the model to spend freely.
OpenAI Cuts GPT-5.6 Sol API Prices After Holding the Line for Months OpenAI cut API pricing on its flagship GPT-5.6 Sol model by more than 20% starting August 21, 2026, the first price cut to its top-tier model since launch. The move follows earlier discounts to its Terra and Luna models and comes as Anthropic and Chinese labs like DeepSeek and Moonshot AI undercut OpenAI on enterprise API pricing. - OpenAI GPT-5.6 Sol API price cuts announced - when did OpenAI reduce GPT-5.6 Sol pricing
That sounds technical, but the business point is plain. You get a dial on the bill.
For a startup or enterprise team running thousands of prompts through a product every day, that matters more than a benchmark chart. A model that is 3% better on a test but twice as expensive in production has to earn that difference in real work. Many won't. The FT's reporting shows customers are already making that calculation, and Anthropic has now built the calculation into the product itself.
Fable 5 is still where the pressure shows #
While Opus 5 got cheaper, Fable 5 became more tightly managed inside Claude subscriptions. Anthropic's help center says that, from July 20, Max plans and premium Team seats can use Fable 5 for up to 50% of their weekly usage limits at no extra cost. It also says that Fable 5 draws from the same regular weekly pool as other models. You don't get a second allowance. You spend the same allowance faster. Pro users and standard Team seats are treated differently. Anthropic says Fable 5 runs on usage credits for those plans, with eligible users receiving a one-time credit to ease the change. Usage-based enterprise plans and API users pay standard API rates. That's not hard to read. Anthropic wants Fable 5 available, but it doesn't want every paid subscriber treating the most expensive model as an unlimited utility.
Frankly, that is the uncomfortable part for a company selling better reasoning as the thing worth paying for. If the best model is too expensive for many customers and too costly to hand out freely, the business has to move in two directions at once. It needs a cheaper model good enough for most work, and it needs strict controls around the model people still want most.
Opus 5 is the first part of that answer. It gives companies a model with near-flagship positioning at a price they can explain internally. Fable 5 is the second part, and it is less flattering: caps and credits, drawn from a shared usage pool, are the mechanics of scarcity. You can call them reliability measures if you like. Customers will still see the meter.
The risk for Anthropic is not that enterprises stop caring about model quality. They care a lot. The risk is that quality gets judged against the invoice every week, not against a launch post. DeepSeek made that lesson impossible to ignore for the whole AI market, and Microsoft, Amazon and Google have the distribution to keep pressing it. Anthropic can still win serious buyers, but Opus 5 shows what the next contract fight looks like: less awe, more arithmetic.
Anthropic's Annualized Revenue Rockets to $65 Billion Ahead of Its IPO Anthropic told investors its annualized revenue run rate hit $65 billion in July, up sevenfold from $9 billion at the end of 2025. Enterprise demand and Claude Code drove the surge, pushing Anthropic past OpenAI's roughly $40 billion pace as both companies race toward IPOs. - Anthropic revenue growth rate before IPO - how fast is Anthropic growing financially
Also read: ShipHero CEO Aaron Rubin Says Claude Mythos Beat His Pentest Firm for $10K • OpenAI's AI Models Broke Out and Hacked Hugging Face, So It Built Daybreak • Vertiv stock loses $12 billion in a week as bond yields rattle AI bets