Hacker News front pageJul 1, 2026, 01:59 UTC+8

Anthropic launches Claude Sonnet 5, closing the agentic gap with Opus 4.8 at a lower price

Claude Sonnet 5

Claude Sonnet 5 is Anthropic's most agentic mid-tier model yet—it plans, uses browsers and terminals, and runs autonomously. Its agentic performance jumps well past Sonnet 4.6 and lands close to Opus 4.8, at $3/$15 per million input/output tokens (introductory $2/$10 through Aug 31, 2026). Safety evals show fewer undesirable behaviors than Sonnet 4.6 and far lower cybersecurity capability than Opus models. Early testers report it finishes multi-step tasks end-to-end without stalling and checks its own output unprompted.

Why it matters: Anthropic's mid-tier workhorse gets a major agentic upgrade with clear pricing — a same-day must-write. Score stays below 90 because the post only shows benchmark comparisons without task completion rates or latency numbers; real-world performance awaits community testing.

Read the original ↗Export Markdown

Take

The reason to click: Sonnet 5 pushes mid-tier agentic performance to a level that used to require Opus-class models. Anthropic claims its planning, browser, and terminal skills land close to Opus 4.8, at roughly 60% of the price. The BrowseComp and OSWorld curves back this up—Sonnet 5 leaves Sonnet 4.6 in the dust and forms a near-continuous cost-performance line with Opus 4.8. Pick Sonnet 5 for cheaper runs, Opus 4.8 when you need max accuracy.

On safety: Sonnet 5 shows fewer undesirable behaviors than Sonnet 4.6 and far lower cybersecurity capability than Opus models. The subtext is “capable agent, not a hacker.” Early testers say it finishes multi-step tasks without stalling and checks its own output unprompted—that “just works” feel matters more for real adoption than benchmark scores.

Where I'd discount: we only have Anthropic's own blog and system card so far, no independent evals. The regular $3/$15 pricing is a step up from Sonnet 4.6, and the $2/$10 intro rate expires August 31—budget for the higher number. If you're already running Opus 4.8 for agent workloads, Sonnet 5 looks like a solid cost-saving downgrade. If you're on Sonnet 4.6, the autonomy jump should be noticeable.

Source: Hacker News front page · anthropic.com
Published: Jul 1, 2026, 01:59 UTC+8 · Jul 1, 2026