Breaking

Anthropic Releases Claude Sonnet 5 as Solo Benchmarks Trail Opus 4.8

July 1, 2026 at 18:16 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic released Claude Sonnet 5, a new model tuned for agentic use, across all plans on June 30, 2026. While it trails the company's higher-end Opus 4.8 on standalone benchmarks, it was designed to run alongside a new Claude Code feature, and developers are debating how it should be judged.

June 30, 2026 · Anthropic

Claude Sonnet 5: Opus-class agentic power at Sonnet-tier pricing

Anthropic's "most agentic Sonnet yet" trails the higher-end Opus 4.8 on many standalone benchmarks — but bets on cost efficiency and an automated Ultracode workflow to win across real-world tasks.

~63.2%

Agentic coding score (some benchmarks)

Token context window (128k output)

Default

Model on Free & Pro plans

API price per 1M tokens — the value case

Output tokens · lower is cheaper

$10

Sonnet 5
intro, then $15

$25

Opus 4.8

At intro pricing, Sonnet 5 costs under half of Opus 4.8 for output — near-Opus agentic performance at Sonnet-tier cost.

The Ultracode auto-workflow

Top Effort tier kicks off a self-driving loop on complex software tasks

Planning

→

Agent generation

→

Execution

→

Self-verification

The case for

Fast, cheap large-codebase refactoring and terminal tasks
A low-cost implementer offloading "grunt work" from top models
Wins across a full workflow even when weaker solo

The critique

Loses to Opus 4.8 on many standalone benchmarks — "still loses"
High / Ultracode settings burn more tokens, raising effective cost
Seen by some as "more expensive and worse than Opus 4.8"

Continue reading

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