Anthropic’s New AI Model Opus 4.8 Beats GPT-5.5 — Here’s What Changed

TLDR

• Anthropic has released Claude Opus 4.8, an upgrade over Opus 4.7 with better coding and reasoning performance
• The model outperforms OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro on several AI benchmarks
• A new effort control setting lets users adjust how deeply the model thinks based on task needs
• Opus 4.8 is four times less likely to let coding errors go unnoticed compared to its predecessor
• Pricing stays the same at $5 per million input tokens and $25 per million output tokens for standard use

---

Anthropic has launched Claude Opus 4.8, its latest flagship AI model, available to users worldwide starting today. The new model is a direct upgrade to Opus 4.7 and brings improvements across coding, reasoning, and honesty.

Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.

Available today at the same price. pic.twitter.com/EufxL7T1kb

— Claude (@claudeai) May 28, 2026

According to Anthropic, Opus 4.8 outperforms OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro on several AI benchmarks. These include agentic coding, agentic financial analysis, and agentic computer use.

Agentic AI refers to systems that can perform tasks with little or no human input. As businesses increasingly adopt AI agents, performance in this area is becoming more important.

New Effort Control Feature

One of the standout additions is a new effort control setting available on Claude.ai and Claude Cowork. It lets users decide how much processing power the model applies to a given task.

For quick, simple tasks, users can dial down the effort to save time and tokens. For complex work, they can increase it to get a deeper result.

Tokens are the units AI services use to measure what goes in and comes out of a model. Every question you ask an AI uses tokens, whether it’s a simple query or a request to write code.

Using less effort also means using fewer tokens, which can lower costs over time for heavy users.

Coding Safety and Enterprise Tools

On the safety side, Opus 4.8 is four times less likely than Opus 4.7 to let coding errors pass without flagging them. That is a meaningful improvement for developers using the model in production environments.

Anthropic is also launching a research preview of dynamic workflows inside Claude Code. This feature can coordinate hundreds of parallel subagents to handle large-scale codebase migrations.

The model is also described as more honest. Anthropic says it will flag uncertainties in its own work and avoid making unsupported claims.

Fast mode pricing has also improved. It now runs three times cheaper to operate than before, though the listed price for users stays at $10 per million input tokens and $50 per million output tokens.

Standard pricing remains unchanged at $5 per million input tokens and $25 per million output tokens.

IPO Race With OpenAI

The Opus 4.8 launch comes as Anthropic is reportedly closing a $30 billion-plus pre-IPO funding round. The round could value the company at over $900 billion.

Anthropic has not confirmed a public listing date, but a 2026 IPO is being reported as a possibility. The move mirrors similar preparations by OpenAI and SpaceX, both of which are also eyeing public markets.

Anthropic was founded by former OpenAI researchers and has grown into one of the leading AI companies in the world.