Sonnet 5 handles agentic tasks and code work at $2/$10 per million tokens until August 31—benchmark-close to Opus 4.8 at lower cost, worth stress-testing against your workloads.
July 3, 2026
Summary
Developers can now run reasoning-heavy agents and tool-use workflows on a mid-tier model that doesn't require Opus pricing. The introductory rate window forces a decision: migrate workloads before the August 31 cutoff or stay locked into higher Opus costs.
Why it matters
Developers can now run reasoning-heavy agents and tool-use workflows on a mid-tier model that doesn't require Opus pricing. The introductory rate window forces a decision: migrate workloads before the August 31 cutoff or stay locked into higher Opus costs.
Implementation verdict
Sonnet 5 replaces Sonnet 4.6 for agentic and coding tasks; Opus 4.8 remains better for maximum accuracy at high reasoning levels. Requires testing your actual inference patterns against published benchmarks—the model now "completes complex tasks where previous Sonnets would stop short," so real-world validation matters. Worth trying now if you're running multi-step agent chains or tool-use loops; lock in introductory pricing before August 31 or absorb a 50% input token cost increase later.
Sources
Dev Signal
Get briefs like this in your inbox — free, every weekday.
100+ sources compressed into one 4-minute read. Ranked, cited, implementation-ready.