By AnthropicLearning Objectives
- Understand where Claude Sonnet 5 sits in the Claude model family
- Explain what "most agentic Sonnet yet" means in practical terms
- Pick the right plan and pricing path for using Sonnet 5
What Is Claude Sonnet 5?
Claude Sonnet 5 is Anthropic's mid-tier Claude model, released on June 30, 2026. In the Claude lineup it sits between the small, fast Claude Haiku 4.5 and the flagship tiers — Claude Opus 4.8 and the public flagship Claude Fable 5. Anthropic positions it as "the most agentic Sonnet model yet."
The headline claim is efficiency: Sonnet 5 approaches Opus 4.8 on cost-performance curves at higher effort levels, while costing far less to run. As one early tester put it, "Claude Sonnet 5 gets more done with less — same output quality, fewer steps to get there." That makes it Anthropic's default recommendation for most everyday and agentic work where Opus-tier capability is not strictly required.
💡Key Concept
"Agentic" means the model can act, not just answer. An agentic model plans a multi-step task, uses tools like browsers and terminals, checks its own work, and adjusts — rather than producing a single response and stopping. Sonnet 5 is tuned to do this reliably with fewer wasted steps.
What It's Good At
Sonnet 5 shows marked improvements over Sonnet 4.6 across the tasks that matter most for day-to-day AI work:
- Autonomous task planning and execution — breaking a goal into steps and carrying them out
- Tool use — driving browsers and terminals as part of a workflow
- Multi-step coding and debugging — writing, running, and fixing code across a task
- Self-verification — checking its own output without being explicitly prompted to
- Lower hallucination and sycophancy than prior Sonnet generations
Anthropic notes one deliberate trade-off: Sonnet 5 has substantially reduced cybersecurity capabilities compared with the Opus and Mythos-class models — a safety choice, since the highest-risk offensive-security capability is reserved for gated, safeguarded releases.
Pricing
- Default model on Claude Free (with usage limits)
- Projects and Artifacts
- Default model
- Higher usage limits
- Priority access
- Through August 31, 2026
- model id claude-sonnet-5
- From September 1, 2026
For most users, Sonnet 5 is now the model you already get by default on the Free and Pro plans — there is nothing to switch on. Developers reach it through the API using the claude-sonnet-5 model id, and it is available inside Claude Code.
Related Tools
- Claude Opus 4.8 — the flagship-tier model Sonnet 5 approaches on cost-performance
- Claude Fable 5 — Anthropic's most capable public model
- Claude Haiku 4.5 — the faster, cheaper tier below Sonnet
- Claude.ai — the consumer interface where Sonnet 5 is the default model
Strengths
- Opus-adjacent quality at a fraction of the cost — approaches Opus 4.8 at higher effort while running far cheaper
- Strong agentic behavior — plans, uses tools, and self-verifies in fewer steps
- Default everywhere — the standard model on Free and Pro, so most users get it automatically
- Lower hallucination and sycophancy than earlier Sonnet generations
Limitations and Considerations
- Not the top of the lineup — for the hardest reasoning, coding, or research, Opus 4.8 and Fable 5 remain more capable
- Reduced cybersecurity capability by design — high-end offensive-security work is reserved for gated Mythos-class releases
- Introductory API pricing is time-limited — the $2 / $10 rate runs only through August 31, 2026, then rises to $3 / $15
Key Takeaways
- Claude Sonnet 5 (released June 30, 2026) is Anthropic's most agentic mid-tier model, approaching Opus 4.8 on cost-performance while running far cheaper
- It plans, uses tools, and self-verifies multi-step tasks in fewer steps, with lower hallucination and sycophancy than prior Sonnet generations
- It is the default model on Claude Free and Pro and is available in Claude Code and via the API (
claude-sonnet-5) - Introductory API pricing is $2 input and $10 output per million tokens through August 31, 2026, then $3 and $15
