Morning Edition · Friday, September 25, 2026Published at 9:31 AM EDT · New York
The company says the model matches Claude Fable 5.1 on most work while running 40 percent cheaper than Opus 5 and generating output more than 30 percent faster.

Anthropic's Claude Opus 5.5 is priced at $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5, with cache-read costs cut by 60 percent. The company says the model matches Claude Fable 5.1's performance on most work at 40 percent lower running cost than Opus 5, and produces output more than 30 percent faster, as reported by The Decoder.
The capability claims are Anthropic's own. The company reports leads on agentic coding, computer use and knowledge work on its internal evaluation suite, and VentureBeat's coverage frames the API price as 60 percent below Fable 5.1 on key agentic tests. Neither the 40 percent running-cost figure nor the agentic performance claims have been reproduced by an independent evaluator yet, and the two cost percentages describe different comparisons: one against Opus 5, the other against Fable 5.1.
Two engineering details matter alongside the pricing. Anthropic says Opus 5.5 is less likely to take hard-to-reverse actions and more resistant to prompt injection than Opus 5, which matters more to agent operators than headline benchmark scores. The model is available directly through Anthropic and on Amazon Web Services, Google Cloud and Microsoft Azure, so switching costs are low for teams already using one of those clouds.
Anthropic, which converts a modest posted price cut into a narrative of parity at lower cost, and cloud resellers on Amazon Web Services, Google Cloud and Microsoft Azure that earn on volume rather than unit price.
The posted prices of $4 and $20 per million tokens and the 60 percent cache-read cut are verifiable, and independently places Opus 5.5 at the top of its intelligence index while reporting somewhat lower figures on Terminal-Bench 4.0 and Humanity's Last Exam, but the headline cost savings are two different comparisons (20 percent off Anthropic's own prior list price, versus a claimed 40 percent lower running cost that depends on Anthropic's own token-efficiency measurements).
Part of a tracked trend
Frontier Model Price War
Frontier API vendors increasingly compete on strategic price-cutting rather than pure capability, repeatedly launching or repricing models below prevailing rates to grab share and compressing industry inference margins; expect recurring below-rival pricing moves.
Start a discussion in Townsquare.
More from this edition
An open-source-intelligence read of how likely this story is true with its real nuance, not a judgment of any outlet. It assesses the claim, weighing independent and adversarial reporting. How we label confidence.
What this means
The mechanism is cost per completed task rather than cost per token. A cheaper, faster model at equal quality increases the number of agent steps a customer can afford per workflow, which increases total token volume even as unit prices fall. Anthropic is exposed on gross margin if the efficiency gain proves smaller than claimed, and OpenAI and Google are exposed on their highest-priced tiers, since a rival's flagship model now undercuts its own predecessor by 20 percent on posted rates. Each of the last several frontier releases across labs has launched at a lower posted price than the model it replaced, so this move continues that pattern rather than breaking from it. Over the coming weeks, a competitor is likely to match this pricing or introduce a cheaper high-end tier, conditional on Anthropic's serving economics holding up under agentic load. That reading would be wrong if independent agentic evaluations place Opus 5.5 materially below Fable 5.1 on long-horizon coding, which would mean the price cut is really a downgrade sold as a discount.
What to watch
Observations to monitor, not financial advice.
Synthesized from: Anthropic · VentureBeat · The Decoder
Comments
0No comments yet.