Claude Models

Claude models are the underlying AI language models made by Anthropic that power Claude Code, Claude.ai, and the broader Claude API platform. They range from smaller, faster models optimised for quick responses to larger, more capable models designed for complex reasoning and deep code analysis.

As of February 2026, the Claude model family comprises three distinct tiers, each suited to a different class of task. Sonnet 4.6 represents a significant capability jump, narrowing the gap between Sonnet and Opus for many workloads.

Model Tiers

Claude Haiku

Haiku is the smallest and fastest model in the Claude family. It is designed for high-throughput, low-latency tasks where speed and cost efficiency matter most.

Best used for:

• Simple code completions and short rewrites
• Quick factual lookups and summarisation
• High-volume, routine tasks where response time is critical
• Drafting and iteration where you need rapid feedback

Claude Sonnet

Sonnet sits in the middle of the model range, offering a strong balance between speed and capability. With the release of Sonnet 4.6 in February 2026, it now approaches Opus-level performance in many domains while maintaining its speed and cost efficiency. It's the default choice for most day-to-day development work and is available on claude.ai and Claude Code.

Sonnet 4.6 improvements:

• Substantial gains in coding, computer use, long-context reasoning, and design work
• 1M token context window (beta)
• 70% of user preference tests favored it over Sonnet 4.5
• In Claude Code testing, it was preferred over Opus 4.5 roughly 59% of the time
• Better resistance to prompt injection attacks

Best used for:

• Feature research and planning
• Writing and reviewing code across a full feature
• Answering technical questions that require complex reasoning
• Computer use tasks (form filling, multi-tab coordination)
• Tasks where you need strong quality without the latency or cost of Opus

Claude Opus

Opus is the largest and most capable model in the Claude family. It excels at tasks that demand deep reasoning, nuanced analysis, and careful multi-step problem solving.

Best used for:

• Complex architectural decisions and design reviews
• Pull request reviews that require subtle judgement about code quality
• Debugging difficult or ambiguous issues
• Tasks where getting the answer right matters more than getting it fast

Choosing the Right Model

When using Claude Code or dispatching agents, you can often specify which model to use. The general rule is straightforward:

In many agent workflows, the model choice is made automatically based on the complexity of the task at hand. Simpler sub-tasks may use Haiku, general work uses Sonnet 4.6, and only the most demanding reasoning tasks require Opus. Sonnet 4.6 is now the default for most Claude Code operations and often outperforms Opus on specific tasks like coding and computer use.

About Anthropic

All Claude models are developed by Anthropic, an AI safety company focused on building reliable, interpretable, and steerable AI systems. Anthropic publishes model documentation and usage guidance on the Claude documentation site.