Breaking

Arena opens a 48-hour Direct Mode window for Claude Sonnet 5.5 High

September 29, 2026 at 19:03 EDT

Foundation Models
AI Agents

Arena will let people test Anthropic’s Claude Sonnet 5.5 at its High effort setting in Direct Mode for 48 hours, starting September 30 at 8 a.m. Pacific Time. After that window, the model stays in Battle Mode and Agent Mode, where it is already listed.

September 30, 2026 · Arena

48 hours with Sonnet 5.5 High

Direct Mode gets a timed listing of Claude Sonnet 5.5 at High effort. Battle Mode and Agent Mode already include it and keep it afterward.

Terminal-Bench 4.0, Anthropic’s scores

Share of agentic coding tasks completed. Column height matches the score.

10.3%Sonnet 5

66.4%Opus 5.5

70.6%Sonnet 5.5

Sonnet 5.5 clears its predecessor by a wide margin and edges the pricier Opus 5.5 on this test.

48h

Direct Mode window from 8 a.m. PT

per million input tokens, half of Opus

30%+

faster output than Sonnet 5

What early users report

Faster reviews, fewer steps, and app builds that needed fewer iterations than Opus 5 in one vendor test.

What independent tests flag

At max effort, token use is the highest measured, and cost per task runs above Sonnet 5 despite the same list price.

How the listing works

Opens in the Direct Mode menu · High effort only for this window · Ends after 48 hours · Battle and Agent Mode keep the model

Continue reading

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