Breaking

Arena Puts Claude Opus 5.5 and GPT-6 Sol Into Side-by-Side Testing

September 23, 2026 at 16:16 EDT

Foundation Models
AI Agents
Software Dev & Coding

Arena has begun showcasing side-by-side outputs from Anthropic's Claude Opus 5.5 and OpenAI's GPT-6 Sol, turning two same-day model launches into a live comparison for users. The first public Arena scores for Claude Opus 5.5 are still pending, so the immediate story is less a settled winner than a new test of how premium agentic models perform on real tasks.

September 23, 2026 - Arena

A Live Test For Two New Workhorse Models

Claude Opus 5.5 and GPT-6 Sol are being compared side by side as Anthropic and OpenAI push frontier models toward lower-cost agentic work.

50%

GPT-6 Sol API price cut vs GPT-5.6 Sol

40%

Claude Opus 5.5 typical workload cost drop vs Opus 5

$10

GPT-6 Sol output price per 1M tokens

$20

Claude Opus 5.5 output price per 1M tokens

Output Token Price

One stacked block equals $2 per 1M output tokens.

$10GPT-6 Sol

$20Claude Opus 5.5

What The Race Is Really About

Ceiling: Opus 5.5 is positioned for difficult coding, audits and long knowledge-work sessions.

Iteration: Sol pushes the price floor lower for repeated agent attempts.

Arena: Opus 5.5 scores are still pending, so live user tasks become the next signal.

Positive Read

Developers get stronger models with lower claimed task costs, more caching value and broader tool support for agent workflows.

Open Questions

The best choice depends on workload, safeguards, latency and whether Arena's pending results match vendor and third-party benchmarks.

Continue reading

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