Anthropic's Claude Sonnet 4.6 Brings Opus Power at Sonnet Price
Anthropic has released Claude Sonnet 4.6 with a 1 million token context window in beta, major coding and computer use improvements, and unchanged pricing — making it the new default across all plans including the free tier.
A New Default for Everyone
Anthropic released Claude Sonnet 4.6 on February 17, 2026, making it the default model for all users — from free-tier subscribers to enterprise API customers. The release maintains Anthropic's roughly four-month update cadence and marks the most significant capability jump for a mid-tier model the company has delivered to date.
The headline feature is a 1 million token context window, currently available in beta through the API. That is double the previous maximum for any Sonnet model and enough to process entire codebases, lengthy legal contracts, or dozens of research papers in a single request, according to Anthropic.
Benchmark Gains That Blur the Tier Lines
Sonnet 4.6 posts a 72.5% score on OSWorld-Verified, a benchmark measuring an AI model's ability to control a real computer interface — navigating spreadsheets, filling multi-step web forms, and managing files across applications. When the computer-use capability first launched in October 2024, Claude scored just 14.9% on the same benchmark, meaning performance has nearly quintupled in roughly 16 months.
On SWE-bench Verified, a widely used software-engineering evaluation, Sonnet 4.6 reached 79.6%, placing it firmly in territory previously occupied only by frontier-class models. Anthropic's own user testing found that developers using Claude Code preferred Sonnet 4.6 over its predecessor Sonnet 4.5 roughly 70% of the time — and preferred it over the older flagship Opus 4.5 59% of the time.
Enterprise validation is already emerging. Box CTO Ben Kus reported that internal testing showed Sonnet 4.6 "outperforming Claude Sonnet 4.5 in heavy reasoning Q&A by 15 percentage points" on enterprise documents, a meaningful gap in production workloads.
Flagship Capability, Mid-Tier Cost
Pricing remains unchanged from Sonnet 4.5: $3 per million input tokens and $15 per million output tokens. By comparison, Anthropic's Opus models carry a $15/$75 price tag — five times higher. Analysts at VentureBeat described the release as a model that "matches flagship AI performance at one-fifth the cost," a dynamic they expect to accelerate enterprise adoption by lowering the financial barrier to deploying capable AI at scale.
New developer capabilities include support for adaptive and extended thinking, context compaction for long sessions, integrated web search with automatic code filtering, and improved instruction-following precision. On the safety side, Anthropic reports major improvements in hallucination resistance and resilience against prompt injection compared to Sonnet 4.5, with safety scores approaching those of Opus 4.6.
Broad Platform Availability
Sonnet 4.6 is available immediately via the Anthropic API, Claude Code, GitHub Copilot, and Amazon Bedrock. The 1 million token context window remains in beta and is currently limited to direct API access — users on the standard interface encounter a smaller context ceiling for now.
The launch follows Anthropic's release of Opus 4.6 on February 5, with an updated Haiku model expected soon, continuing the company's staggered portfolio refresh strategy.
What It Means for the AI Landscape
The release underscores a broader industry pattern: capabilities that once required the most expensive frontier models are migrating down the pricing ladder at speed. For businesses that delayed AI deployment due to cost concerns, Sonnet 4.6 removes a significant obstacle. For developers building autonomous agents, the combination of a million-token context window and strong agentic planning scores creates new architectural possibilities that were impractical just months ago.
