Claude Sonnet 4.6: Free? Opus Access? Claude Code Setup & Every Model Compared

The Claude 4 model family — currently led by Sonnet 4.6 and Opus 4.6 — is Anthropic's answer to a simple problem: the old Opus/Sonnet/Haiku gap was too expensive to justify. Anthropic released Sonnet 4.6 on February 17, 2026, just 12 days after Opus 4.6. "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," Anthropic said in a blog post. The headline result: Sonnet 4.6 scores 72.5% on OSWorld-Verified for computer use — within 0.2% of Opus 4.6 — and 79.6% on SWE-bench Verified for coding, leading all models in office productivity tasks at 1633 Elo on GDPval-AA.

For most users the practical upshot is simple: for the first time, a Sonnet model is preferred over the previous generation's Opus in coding evaluations, with computer use accuracy hitting 94% on insurance benchmarks. Combined with Sonnet 4.6 being the free-tier default, most people no longer need a paid plan for capable AI. This guide covers the full Claude 4.x family, what changed in each release, who gets Opus 4.6 access, how to use specific models in Claude Code, full API pricing, and the Gemini 2.5 Pro comparison everyone is searching for.

Claude 4.x Model Family: Full Release Timeline

Model	Released	Key Additions	API Model String
Claude Haiku 4.5	Mid-2025	Fastest/cheapest model; near-Sonnet 4 performance; targets high-volume and real-time apps	claude-haiku-4-5-20251001
Claude Sonnet 4.5	Mid-2025	Complex agents, coding, extended thinking; highest intelligence across most tasks at launch	claude-sonnet-4-5-20250514
Claude Opus 4.5	November 24, 2025	Coding and workplace tasks (spreadsheets); Infinite Chats feature eliminating context window errors; 67% price cut vs. Opus 4	claude-opus-4-5
Claude Opus 4.6	February 5, 2026	Agent Teams (multi-agent coordination), 1M token context window (beta), Claude in PowerPoint, longest METR task horizon (14.5 hrs)	claude-opus-4-6
Claude Sonnet 4.6	February 17, 2026	Upgraded coding, computer use, long-context reasoning, agent planning, knowledge work, design; same price as 4.5; 1M context beta; default across all plans	claude-sonnet-4-6

What Is Claude Sonnet 4.6? What's New?

Sonnet 4.6 is a full upgrade across the model's skills in coding, computer use, long-context reasoning, agent planning, knowledge work, and design, featuring a 1M token context window in beta. The most striking benchmark result: ARC-AGI-2 novel problem-solving jumps from 13.6% (Sonnet 4.5) to 58.3% (Sonnet 4.6) — a 4.3x improvement and the largest single-generation gain recorded.

Sonnet 4.6 has a reliable knowledge cutoff of August 2025, compared to Opus 4.6's May 2025 and Haiku 4.5's February 2025. Both Opus and Sonnet default to 200,000 max input tokens but can stretch to 1 million in beta at a higher cost.

Sonnet 4.6 Benchmark Results at a Glance:

• Coding (SWE-bench Verified): 79.6% — up from 77.2% on Sonnet 4.5, approaching Opus 4.6's 80.8%
• Computer use (OSWorld-Verified): 72.5% — within 0.2% of Opus 4.6 (72.7%)
• Office productivity (GDPval-AA): 1633 Elo — leads all models including Opus 4.6
• Financial tasks (Finance Agent): 63.3% — best-in-class
• Scaled tool use (MCP-Atlas): 61.3% — beats Opus 4.6's 60.3%
• Novel reasoning (ARC-AGI-2): 58.3% — up from 13.6% on Sonnet 4.5

What Is Claude Opus 4.6? The Flagship Explained

Anthropic released Opus 4.6 on February 5, 2026. The main improvements included an Agent Team feature and Claude in PowerPoint. As of February 20, 2026, it is the model with the longest task-completion time horizon as estimated by METR, having a 50%-time horizon of 14 hours and 30 minutes and an 80%-time horizon of 1 hour 3 minutes.

Opus 4.6's defining capability over Sonnet 4.6 is agentic depth: the 1M token context window and native multi-agent collaboration open new categories of work that weren't possible before — large codebase analysis, complex architecture decisions, and agent teams where Claude instances coordinate with each other to complete long-horizon tasks. In one notable demonstration, 16 Claude Opus 4.6 agents wrote a C compiler in Rust from scratch, capable of compiling the Linux kernel — the first model able to accomplish this. The experiment cost nearly $20,000.

What Is Claude Haiku 4.5? (And Is It Good?)

Claude Haiku 4.5 is Anthropic's fastest and most cost-efficient model with near-frontier performance, ideal for real-time applications, high-volume processing, and cost-sensitive deployments requiring strong reasoning. It matches Claude Sonnet 4's (not 4.5's) performance on coding, computer use, and agent tasks. Claude in Chrome now defaults to Haiku 4.5 so it's a faster, more responsive experience — users can always switch back to Sonnet 4.5.

Haiku 4.5 is not weaker across the board — for tasks like classification, routing, simple Q&A, file reads, and quick edits, it's the right tool. Its knowledge cutoff is February 2025, six months older than Sonnet 4.6's August 2025 cutoff, which matters for code frameworks and current events.

Is Claude Sonnet 4.6 Free? Who Gets What Model

Plan	Price	Models Available	Claude Code?
Free	$0	Sonnet 4.6 (rate-limited, default)	❌ No
Pro	$20/month	Sonnet 4.6 + Opus 4.6 + Haiku 4.5; 5x usage vs. Free; Research, Google Workspace, extended thinking	✅ Yes
Max 5x	$100/month	All models incl. Opus 4.6 with 1M context + Agent Teams; 5x Pro limits; adaptive thinking	✅ Full access
Max 20x	$200/month	All models; 20x Pro limits; zero-latency priority; early access to new features	✅ Full access
Team	$25/seat/month	Sonnet 4.6 + Opus 4.6; shared workspaces; Claude Code included with every standard seat	✅ Included
Enterprise	Custom	All models; SSO, data privacy guarantees, dedicated support; now self-serve at claude.ai/create/enterprise	✅ Included

Is Opus 4.5 / 4.6 on Claude Code? Full Claude Code Model Guide

Anthropic recommends upgrading to Claude Sonnet 4.6 and Claude Haiku 4.5 respectively, as Claude Sonnet 3.7 and Claude Haiku 3.5 have been retired. Sonnet 4.6 delivers improved agentic search performance while consuming fewer tokens.

Claude Code Model Availability by Plan:

• Free plan: Claude Code not available — requires at least a Pro subscription or API credits
• Pro plan ($20/month): Claude Code included; default model is Sonnet 4.6; can use Opus 4.6 with Pro's message limits
• Max plans ($100–$200/month): Full Claude Code access; Opus 4.6 with higher limits and Agent Teams
• API (pay-per-token): Claude Code CLI with any model — no monthly cap, charged per token

How to Switch Models in Claude Code:

In the Claude Code CLI, use the /model command to switch between models during a session:

• /model claude-sonnet-4-6 — switch to Sonnet 4.6 (recommended for most coding tasks)
• /model claude-opus-4-6 — switch to Opus 4.6 (for complex multi-file reasoning, architecture decisions)
• /model claude-haiku-4-5-20251001 — switch to Haiku 4.5 (fast reads, simple edits, 1/3 Sonnet cost)
• /compact — manually trigger context compaction to manage long session costs

Claude Code automatically selects models based on task complexity in smart routing mode — Haiku 4.5 for simple file reads, Sonnet 4.6 for standard coding tasks, and Opus 4.6 for state-of-the-art software engineering tasks. CLAUDE.md and system prompts are automatically cached. After the first request, cached content costs only 10% of the original price.

Claude API Pricing: Full 2026 Breakdown

Model	Input /MTok	Output /MTok	Long Context (>200K input)	Max Context
Claude Haiku 4.5	$1.00	$5.00	Standard only	200K tokens
Claude Sonnet 4.6	$3.00	$15.00	$6.00 input / $22.50 output	1M tokens (beta)
Claude Opus 4.6	$5.00	$25.00	2x pricing above 200K	1M tokens (beta)
Claude Opus 4.6 Fast Mode	$30.00	$150.00	No extra long-context charge	1M tokens (full)
Any model (Batch API — 50% off)	Half standard rate	Half standard rate	Discount applies	Same as standard
Prompt cache reads (any model)	10% of base rate	Standard output rate	Discount stacks	—

The 200K Token Pricing Trap (Read This Before Sending Large Contexts):

Sonnet 4.6 and Opus 4.6 have a pricing cliff: if your input is 200K tokens or under, you pay $3/MTok. If your input crosses 200K tokens by even one token, the entire request charges at $6/MTok input and $22.50/MTok output — double the normal input rate. For large document or codebase analysis, consider chunking requests under 200K tokens or enabling the 1M context beta (which prices transparently by default at the long-context rate). Opus 4.6 Fast Mode sidesteps this — it includes the full 1M context at no additional long-context charge.

API Cost-Saving Stack:

• Prompt caching: 90% off cached input tokens — cache your system prompts and long context blocks. On 10 requests to the same 100K token system prompt, you pay ~$0.30/MTok reads vs. $3.00/MTok full price
• Batch API: 50% off all models for non-real-time workloads — results within 24 hours. Sonnet 4.6 drops to $1.50/$7.50 per MTok
• Model routing: Use Haiku 4.5 for classification and simple tasks ($1/MTok), Sonnet 4.6 for most work ($3/MTok), Opus 4.6 only for complex reasoning ($5/MTok). Routing correctly can cut API spend 60–80%
• US-only inference: Adding the inference_geo: "us" parameter routes to US-only servers at 1.1x pricing — useful for compliance, not cost savings

Does Claude 4.6 Support Images?

Yes — all current Claude models (Haiku 4.5, Sonnet 4.6, Opus 4.6) support image inputs via the API and on claude.ai. You can upload JPEGs, PNGs, GIFs, and WebP images. Claude can analyze charts, read documents in image form, interpret UI screenshots, describe photos, and reason about visual content. Image tokens are counted in your input token usage — a typical image uses approximately 1,000–1,500 tokens depending on resolution. Claude handles image uploads for you: give Claude an image and tell it where to upload, whether it's an expense report, form attachment, or a picture upload.

How to Create a Document Using Claude

Claude can create, format, and export documents directly from conversations. The most reliable approach for structured documents:

1. Start your message with the output format: "Write me a Word document / Markdown report / Google Doc outline that..."
2. Claude generates the content as an Artifact — a separate panel you can edit and download
3. For file creation in Claude.ai desktop or Claude Code: explicitly request "save this as a .md / .docx file" — Claude writes and presents the file for download
4. In Claude Code CLI: Claude can create files directly in your filesystem — "create a new file called report.md and write..."
5. Via Claude in Excel or Claude in PowerPoint (Opus 4.6 powered): Claude can build and edit spreadsheets and slide decks natively within the application

Is Gemini 2.5 Pro Better Than Claude Sonnet 4.6?

This is the most-searched comparative question about the Claude family in early 2026, and the answer depends on what you're measuring:

Task	Claude Sonnet 4.6	Gemini 2.5 Pro	Edge
Coding (SWE-bench)	79.6%	~63% (est.)	Claude Sonnet 4.6
Computer use	72.5% OSWorld	Not natively supported	Claude Sonnet 4.6
Long context	1M tokens (beta)	1M tokens	Roughly equal
Multimodal (video/audio)	Image/text	Video, audio, image, text	Gemini 2.5 Pro
Office productivity	1633 Elo (best-in-class)	Lower on GDPval-AA	Claude Sonnet 4.6
Free tier quality	Sonnet 4.6 (strong)	Gemini 2.5 Flash (strong)	Roughly equal
API pricing (comparable tier)	$3/$15 per MTok	Competitive, slightly lower	Gemini slight edge

Summary: Claude Sonnet 4.6 is the stronger choice for coding, computer use, and office automation tasks. Gemini 2.5 Pro holds an advantage for native video and audio understanding. For pure text reasoning and complex agentic tasks, Claude Opus 4.6 is the current benchmark leader. Neither is universally better — the right choice depends on your use case.

Is Claude 4.6 Better Than ChatGPT / GPT-5?

Claude Sonnet 4.6 vs. GPT-5.2 is the most direct comparison for coding and knowledge work. On SWE-bench Verified, Sonnet 4.6 scores 79.6% vs. GPT-5.2's 74.9%. On office productivity tasks (GDPval-AA), Sonnet 4.6 leads all models. GPT-5.2 Pro (extended reasoning mode) edges ahead on deep reasoning-heavy benchmarks like GPQA Diamond and Humanity's Last Exam. Claude Opus 4.6 maintains a lead on the most demanding agentic tasks and long-horizon task completion. At comparable pricing tiers, both are extremely competitive — practical performance differences are often negligible for everyday use cases, with gaps emerging primarily in specialized benchmarks.

Frequently Asked Questions

Is Claude Sonnet 4.6 Free?

Yes — Sonnet 4.6 is the default model for Claude Free plan users. The free tier now includes file creation, connectors, skills, and context compaction. Free users face rate limits during high-traffic periods and do not get Claude Code access, but Sonnet 4.6 itself is fully available at no cost.

What Is Claude Haiku 4.5?

Haiku 4.5 is Anthropic's fastest, cheapest current-generation model — positioned for real-time applications, high-volume processing, and cost-sensitive tasks. It matches Sonnet 4 (not 4.5) performance on coding and agent tasks. API pricing is $1/$5 per million tokens. Knowledge cutoff is February 2025. Available on Pro, Max, Team, and Enterprise plans; and in Claude in Chrome as the default model for browser tasks.

Is Opus 4.5 / 4.6 on Claude Code?

Yes. Claude Max $100 and $200 plans provide access to Claude Opus 4.6 with 1M context window, Agent Teams, and adaptive thinking in Claude Code — handling professional development with large projects, extended coding sessions, and state-of-the-art software engineering. Pro plan users can access Opus 4.6 in Claude Code with standard 200K context limits. Free plan users cannot access Claude Code at all.

How to Use Opus 4.6 in Claude Code for Free?

There's no free path to Opus 4.6 in Claude Code — it requires at minimum a Pro plan ($20/month). The closest free alternative: use the Anthropic API with pay-as-you-go billing (no monthly subscription, charged per token at $5/$25 per million tokens for Opus 4.6). For light exploration, this can cost under $5 before you need to seriously budget API spend.

How to Swap Claude Code to Sonnet 4.6?

In the Claude Code CLI, run /model claude-sonnet-4-6 during any session to switch. To set it as default, configure your Claude Code settings file: model: "claude-sonnet-4-6". Alternatively, pass the model string directly in your API call: "model": "claude-sonnet-4-6".

What Is Claude Haiku 4.5 Good For?

Haiku 4.5 excels at tasks where speed and cost matter more than maximum reasoning depth: customer service chatbots, document classification, quick code edits and file reads, content moderation, real-time autocomplete, and any high-volume workload where you're making thousands of API calls. At $1/$5 per MTok, it's one-third the cost of Sonnet 4.6 and five times cheaper than Opus 4.6.

Is Claude Sonnet 4.5 Better Than Claude 4 (Sonnet 4)?

Yes across all core benchmarks. Sonnet 4.5 improved on Sonnet 4 in coding (SWE-bench: 77.2% vs. ~56%), computer use (OSWorld: 61.4% vs. 42.2%), and agentic task performance. Sonnet 4.6 is a further improvement over 4.5 on every major benchmark, making it the recommended model for all new work.

How Much Does Claude Sonnet 4.6 Cost?

Via claude.ai: free (with limits) on the Free plan; included in Pro ($20/month) and all higher plans. Via API: $3.00 per million input tokens and $15.00 per million output tokens for standard context requests (up to 200K input tokens). For long-context requests exceeding 200K tokens, input doubles to $6.00/MTok and output to $22.50/MTok. Batch API discount: 50% off, bringing standard rate to $1.50/$7.50.

When Was Claude Sonnet 4.5 Released?

Claude Sonnet 4.5 was released in mid-2025 (May–June 2025 timeframe) as part of the Claude 4 generation launch. It has since been superseded by Claude Sonnet 4.6 (February 17, 2026) as the default model on claude.ai. Sonnet 4.5 remains available via API using the model string claude-sonnet-4-5-20250514.