Overview
Claude Sonnet 4.6 is Anthropic's most capable Sonnet model yet — a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. Released on February 17, 2026, it arrived just 12 days after Claude Opus 4.6, signaling Anthropic's aggressive push in the AI race.
What makes this launch remarkable is the value equation. Claude Sonnet 4.6 is a mid-range AI model that works as well as top-of-the-line models like Opus 4.6, but costs only one-fifth as much. For developers, enterprises, and everyday users, that is a game-changer.
This article covers every major upgrade, benchmark result, pricing detail, and practical use case — everything you need to know about Claude Sonnet 4.6.
What's New in Claude Sonnet 4.6 at a Glance
| Feature | Claude Sonnet 4.5 | Claude Sonnet 4.6 |
|---|---|---|
| Context Window | 500K tokens | 1M tokens (beta) |
| Pricing (Input/Output) | $3 / $15 per 1M tokens | $3 / $15 per 1M tokens |
| Default for Free/Pro | No | Yes |
| OSWorld Score (Computer Use) | ~61.4% | ~72.5% |
| Adaptive Thinking | Limited | Yes |
| Extended Thinking | Limited | Yes |
| Compaction (Beta) | No | Yes |
| Claude Cowork Default | No | Yes |
The 1 Million Token Context Window
One of the biggest headlines from this release is the context window. The beta version of Sonnet 4.6 features a 1 million token context window — double the size of the previous largest window available for Sonnet models.
To put that in plain terms: the 1M token context window is enough to hold entire codebases, lengthy contracts, or dozens of research papers in a single request.
For developers using Claude Code, this is transformative. With 1 million tokens, Sonnet 4.6 can ingest entire codebases — even extremely large ones — by seeing the entirety of extremely large horizons at once, allowing it to understand the full scope of dependencies at once and follow flow paths at longer depths.
For enterprise users, the benefit is equally powerful. Teams working with contracts, research documents, or financial reports can load all relevant materials into a single session and have the model reason across all of them simultaneously — no more re-uploading, no more lost context.
Coding: The Biggest Leap Forward
Coding is where Claude Sonnet 4.6 truly shines. In Claude Code, early testing found that users preferred Sonnet 4.6 over Sonnet 4.5 roughly 70% of the time. Users reported that it more effectively read the context before modifying code and consolidated shared logic rather than duplicating it, making it less frustrating to use over long sessions than earlier models.
What's even more striking: users preferred Sonnet 4.6 to Opus 4.5 — Anthropic's frontier model from November 2025 — 59% of the time. They rated Sonnet 4.6 as significantly less prone to overengineering and "laziness," and meaningfully better at instruction following. They reported fewer false claims of success, fewer hallucinations, and more consistent follow-through on multi-step tasks.
What Developers Are Saying
| Company | Verdict |
|---|---|
| GitHub | "Already excelling at complex code fixes, especially when searching across large codebases is essential." |
| Cursor | "A notable improvement over Sonnet 4.5 across the board, including long-horizon tasks and more difficult problems." |
| Replit | "The performance-to-cost ratio is extraordinary — it outperforms on orchestration evals and handles our most complex agentic workloads." |
| Cognition | "Has meaningfully closed the gap with Opus on bug detection, letting us run more reviewers in parallel." |
| Windsurf | "For the first time, Sonnet brings frontier-level reasoning in a smaller and more cost-effective form factor." |
Computer Use: From Experimental to Near-Human
When Anthropic first introduced general-purpose computer use back in October 2024, it described the feature as experimental and error-prone. Sonnet 4.6 marks a dramatic leap forward.
In benchmarks designed to test how well AI can navigate web and desktop apps (OSWorld), Claude Sonnet 4.6 scored 72.5%, a significant jump from the 61.4% of its predecessor.
Beyond benchmarks, early Sonnet 4.6 users are seeing human-level capability in tasks like navigating a complex spreadsheet or filling out a multi-step web form, before pulling it all together across multiple browser tabs.
OSWorld Benchmark: Sonnet Progress Over Time
| Model | OSWorld Score |
|---|---|
| Claude Sonnet 3.5 (Oct 2024) | ~22% |
| Claude Sonnet 3.7 | ~50% |
| Claude Sonnet 4.5 | ~61.4% |
| Claude Sonnet 4.6 | ~72.5% |
This means the model can interact with legacy enterprise software — tools built before APIs existed — without any custom connectors. Almost every organization has software it can't easily automate. To have AI use such software, users would previously have had to build bespoke connectors. But a model that can use a computer the way a person does changes that equation.
Security: Improved Prompt Injection Resistance
Computer use also carries risks. Malicious actors can attempt to hijack the model by hiding instructions on websites in what's known as a prompt injection attack. Anthropic's safety evaluations show that Sonnet 4.6 is a major improvement compared to its predecessor, Sonnet 4.5, and performs similarly to Opus 4.6 on prompt injection resistance.
Long-Horizon Reasoning: The Vending-Bench Test
One of the most fascinating capability demonstrations came from the Vending-Bench Arena evaluation, which tests how well a model can run a simulated business over time — including direct competition between AI models.
Sonnet 4.6 developed an interesting new strategy: it invested heavily in capacity for the first ten simulated months, spending significantly more than its competitors, and then pivoted sharply to focus on profitability in the final stretch. The timing of this pivot helped it finish well ahead of the competition.
It ended the simulation with more than double the balance of previous models, proving it can plan for months, not just minutes.
This kind of strategic, multi-step reasoning is exactly what enterprises need for agentic automation — AI that doesn't just respond to prompts but plans and executes over long time horizons.
Pricing: Same Cost, Far More Power
One of the most commercially important aspects of this release is what did NOT change: the price.
| Tier | Price (Per 1M Tokens) |
|---|---|
| Input tokens | $3.00 |
| Output tokens | $15.00 |
| Compared to Sonnet 4.5 | Same pricing |
To put that in perspective, running a high-performance AI agent is now roughly five times cheaper than it was just a few weeks ago compared to running equivalent Opus-class tasks. For enterprises running millions of automated tasks a day, this is a significant cost reduction.
Who Gets Claude Sonnet 4.6 by Default?
Claude Sonnet 4.6 is the default for Free tier users, and Anthropic now includes file creation, connectors, skills, and compaction for free tier users.
| Plan | Default Model | Notes |
|---|---|---|
| Free | Claude Sonnet 4.6 | Includes file creation, connectors, skills, compaction |
| Pro | Claude Sonnet 4.6 | Full feature access |
| API / Enterprise | Selectable | Model string: claude-sonnet-4-6 |
This is a significant upgrade for free tier users, who previously had access to older or less capable models by default.
Where Is Claude Sonnet 4.6 Available?
Claude Sonnet 4.6 rolled out across Anthropic's full ecosystem on February 17, 2026.
| Platform | Status |
|---|---|
| claude.ai (web/mobile/desktop) | Available — default for Free & Pro |
| Claude Cowork (macOS) | Available — default model |
| Anthropic API | Available |
| Amazon Bedrock | Available |
| GitHub Copilot | Rolling out — available in GitHub Copilot |
| Claude Code (CLI) | Available |
New API Features in Sonnet 4.6
Developers using the Anthropic API get access to expanded features with this release:
Within the API, Sonnet 4.6 now supports both adaptive and extended thinking, as well as compaction in beta. That allows users to quickly select optimized features for cost-to-performance and continual execution, even when the context fills up.
| API Feature | Description |
|---|---|
| Adaptive Thinking | Dynamically adjusts reasoning depth based on task complexity |
| Extended Thinking | Enables deeper, step-by-step reasoning for hard problems |
| Compaction (Beta) | Keeps long tasks running even when context window fills up |
| 1M Token Context | Available in beta for large-scale document and code tasks |
Safety and Character
Anthropic conducted extensive safety evaluations before releasing Sonnet 4.6. Safety researchers concluded that Sonnet 4.6 has "a broadly warm, honest, prosocial, and at times funny character, very strong safety behaviors, and no signs of major concerns around high-stakes forms of misalignment."
The model was found to be as safe as, or safer than, Anthropic's other recent models. Its improved resistance to prompt injection attacks is especially important for enterprise computer use deployments.
Design and Visual Work
Beyond coding and reasoning, customers noted a surprising improvement in visual and design outputs. Customers independently described visual outputs from Sonnet 4.6 as notably more polished, with better layouts, animations, and design sensibility than those from previous models. Customers also needed fewer rounds of iteration to reach production-quality results.
This makes Sonnet 4.6 a strong candidate for teams building front-end applications, generating UI mockups, or producing data visualizations.
How Claude Sonnet 4.6 Fits in the Claude Model Family
| Model | Best For | Relative Cost |
|---|---|---|
| Claude Haiku 4.5 | Fast, lightweight tasks, high-volume API calls | Lowest |
| Claude Sonnet 4.6 | Most tasks — coding, reasoning, computer use, knowledge work | Mid (~1/5 of Opus) |
| Claude Opus 4.6 | The hardest, most complex tasks requiring maximum intelligence | Highest |
For most users and most use cases, Sonnet 4.6 is now the right starting point. It closes the gap with Opus to a degree that makes choosing Opus a more deliberate, narrower decision.
Conclusion
Claude Sonnet 4.6 is the most impactful model release Anthropic has made for everyday users and developers. It brings near-Opus intelligence at Sonnet pricing, doubles the context window, dramatically improves computer use, and ships stronger coding capabilities to everyone — including free users.
Performance that would have previously required reaching for an Opus-class model — including on real-world, economically valuable office tasks — is now available with Sonnet 4.6.
If you haven't tried it yet, now is the time. Claude Sonnet 4.6 is live on claude.ai, the API, GitHub Copilot, Amazon Bedrock, and Claude Cowork — available to free users by default, starting today.