Breaking

Anthropic Launches Claude Sonnet 5, Offering Near-Opus 4.8 Power at Lower Cost

July 2, 2026 at 00:45 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic released Claude Sonnet 5, the newest model in its Sonnet line, around June 30, 2026, positioning it as delivering performance close to the higher-end Opus 4.8 at a lower price. Early users praise its cost efficiency, while many still find Opus 4.8 superior on complex visual builds.

June 30, 2026 · Anthropic

Claude Sonnet 5 arrives cheaper and more agentic — but Opus 4.8 keeps the crown

Anthropic's newest model posts big gains in agentic search, tool use and computer control, and lands at a lower price. Yet hands-on build tests still favour Opus 4.8 for stability and real coding — putting the benchmark-vs-practice gap at the heart of the debate.

40%↓

Cheaper input at introductory rate vs Opus 4.8 ($2 vs $5)

Effort levels (medium / high / extra high) to tune cost vs performance

1.35×

New tokenizer can raise token use — eroding the price gap

Price per 1M output tokens

Lower columns = cheaper. Sonnet 5 undercuts Opus 4.8 on both intro and standard rates.

$10

Sonnet 5
intro

$15

Sonnet 5
standard

$25

Opus 4.8

Where it wins

Big agentic gains over Sonnet 4.6
Stronger task completion, tool use, debugging
Leads on BrowseComp & OSWorld-Verified
Faster, cheaper visual generation for some

Where it stumbles

Opus 4.8 more stable in real builds
Unstable orbit/galaxy simulation test
Heavy tokens can match or exceed Opus cost
Hallucinations & missed hardware limits

The takeaway

"Overall Opus 4.8 is stronger, and I won't switch for now."

The gap between benchmark scores and practical performance is the crux of the debate — with rival GLM 5.2 also reported to outperform in real builds. For now, practitioners recommend flexible, per-task model switching rather than a single default.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.