Anthropic Releases Claude Sonnet 5: Near-Opus Performance at Sonnet Pricing

Summary: Anthropic's Claude Sonnet 5, released June 30, narrows the gap with Opus 4.8 and became the default model for all Free and Pro Claude users on July 1.

Key Facts

• Pricing: $2/M input tokens, $10/M output tokens (introductory rate through August 31) — lower than Sonnet 4.6 at launch
• Context: 1 million token context window, up to 128,000 output tokens
• Agentic capability: Completes complex multi-step tasks involving browser and terminal tool use that Sonnet 4.6 could not finish; self-reviews its own output unprompted
• Safety: Improved resistance to prompt-injection hijacks, lower hallucination and sycophancy rates, Adaptive Thinking enabled by default

Why It Matters

Developers who previously needed Opus-tier pricing for high-quality agent workflows can now achieve comparable results at Sonnet prices. For API-driven startups and agent frameworks where per-call cost is a primary constraint, this release meaningfully changes the economics of production-grade AI agents.

Read More

• Introducing Claude Sonnet 5 — Anthropic
• Anthropic launches Claude Sonnet 5 as a cheaper way to run agents — TechCrunch