Anthropic has launched Claude Sonnet 4.6, a model that now delivers performance the company previously reserved for its most expensive tier - at the same price as its predecessor, starting at US$3 per million input tokens and $15 per million output tokens.
The launch comes as enterprise software stocks sit in a prolonged rout, with the S&P 500 Software Index down roughly 20% year-to-date while the broader market is up 9% across the same stretch.
Find out more: SaaSpocalypse continues; software sector down another 3%
Investors are picking apart the long-term viability of per-seat SaaS pricing against AI disruption, and Sonnet 4.6 doesn't exactly ease those concerns.
The upgrades that matter
The main improvements are in computer use, coding, and long-context reasoning.
Anthropic's Sonnet line has posted consistent gains on the OSWorld benchmark across 16 months - a test that runs AI models through real software tasks in simulated environments, no special APIs or connectors involved.
Early users are reporting human-level performance on tasks like navigating complex spreadsheets and completing multi-step web forms.
That puts Sonnet 4.6 in territory that previously required Anthropic's flagship Opus tier.
In Claude Code, users preferred Sonnet 4.6 over its predecessor roughly 70% of the time, and over Opus 4.5 - Anthropic's frontier model from November 2025 - 59% of the time.
Fewer hallucinations, better instruction following, and less overengineering across long sessions were the recurring themes.
The model ships with a 1M token context window in beta, enough to hold entire codebases or lengthy contracts in a single request.
Where the SaaS overlap gets uncomfortable
The connection to the broader software selloff isn't hard to draw.
The bear case running through the sector sits on two tracks: AI agents handling enterprise workflows at lower cost, and AI coding platforms enabling companies to build their own tools rather than licence third-party software.
Sonnet 4.6 moves both scenarios further along.
Its computer use capability is built to interact with legacy systems that pre-date modern API integration - the kind enterprises have historically been locked into, and that SaaS vendors have relied on for retention.
A model that navigates those systems without custom connectors chips away at that lock-in argument considerably.
Anthropic's Excel add-in now pulls live data from S&P Global, LSEG, PitchBook, Moody's and FactSet directly into spreadsheets, sitting on top of workflows that dedicated financial data platforms have long owned.
Databricks reported Sonnet 4.6 matches Opus 4.6 on document comprehension across charts, PDFs and tables, while Box saw a 15 percentage point improvement over Sonnet 4.5 on heavy reasoning tasks.
Insurance workflow firm Pace clocked 94% accuracy on its sector benchmark, which CEO Jamie Cuffe described as “the highest-performing model we've tested for computer use.”
The counter
The pushback from Wall Street has been consistent throughout the selloff.
Wedbush analyst Dan Ives argued the rout was pricing in an "Armageddon scenario" far removed from reality, given enterprises won't dismantle tens of billions in existing software infrastructure overnight.
JPMorgan called it an overly bearish read on AI disruption and flagged short interest at record levels - a positioning imbalance that has typically preceded rebounds in past selloffs.
Anthropic isn't pitching Sonnet 4.6 as a wholesale replacement for enterprise stacks, with the model integrating across GitHub, Cursor, Replit and Zapier as infrastructure running underneath existing tools.
That framing has limits, though.
A model that six months ago required Opus-tier spend now runs at Sonnet prices, and the rate of cost compression in this space shows no sign of easing.
Sonnet 4.6 is available now across all Claude plans, the API and major cloud platforms, with the free tier upgraded to run it by default.
