Anthropic Permanently Locks in Low Pricing for Sonnet 5, Scraps Planned 50% Price Hike
Miles Bennett
Anthropic cancelled Claude Sonnet 5's planned 50% price increase, permanently locking input/output token pricing at $2/$10 per million — a clear scale-over-margin bet on the eve of its IPO.
What was the planned hike, and where does pricing stand now?
Sonnet 5 launched in June at $2 per million input tokens and $10 per million output tokens.
The original plan raised those to $3 and $15 on August 31 — a 50% increase, now permanently cancelled.
This means → enterprises already building workflows on Sonnet 5 will not face a mid-stream cost reset.
Why do agentic AI developers care the most?
Anthropic positions Sonnet 5 as its "most agentic Sonnet model" — it can plan autonomously and use tools like browsers and terminals to complete complex tasks.
Agentic tasks — where AI acts like an assistant executing multi-step operations — typically require many rounds of model calls; token costs compound fast with task complexity.
In plain terms = a single agentic task might call the model dozens of times. A 50% unit-price hike means the total bill rises by far more than 50%. Locked pricing lets enterprises commit to scaling agentic workflows.
Where is Anthropic's money coming from — and going?
Anthropic confidentially filed its IPO registration with the SEC on June 1; the market expects a listing later this year.
In May, it closed a $65 billion Series H at a $965 billion post-money valuation; annualised revenue already exceeds $47 billion, with over 1,000 enterprise clients spending more than $1 million a year.
Compute spending is scaling in parallel: a nearly $45 billion compute-resource deal with SpaceX; Broadcom, Apollo, and BlackRock launched AI XPV Platform with $35 billion in first-tranche capacity to support AI infrastructure including Anthropic.
What is the business logic behind locking prices low?
Revenue is growing fast and compute investment keeps expanding — yet Anthropic chose to freeze prices rather than raise them. The strategy points to using low API pricing to drive adoption scale.
This means → Anthropic's priority is "lock enterprise customers into the ecosystem first," not "maximise per-unit margin now."
This reflects an agentic AI market still in its land-grab phase, where scale moats matter more than near-term margins.
What comes next?
Reports indicate Anthropic is about to release Claude Sonnet 5.5, internal codename "Fennec," with a rumoured context window doubled to 2 million tokens, faster inference, and stronger agentic capability.
Sonnet 5.5's overall performance reportedly approaches flagship-tier Fable 5, but pricing stays at the Sonnet tier.
In plain terms = flagship-class performance at mid-tier pricing — whether this combination can truly lock in enterprise customers is the key test of Anthropic's commercialisation thesis.
Content is for reference only, not financial advice.