AI News: Claude Opus 5.5 — Smarter, Shorter, Faster
Anthropic has released a new flagship model. The biggest difference is not found in benchmarks, but in how it writes.
Figure 01.0 — AI News
First, the terms. Claude is the AI assistant from Anthropic. There are different models: the “brains” behind the assistant. Opus is the most robust line, designed for complex tasks. Fable is Anthropic’s even more powerful flagship model, but more expensive. A benchmark is a standardised test used to compare models with each other.
What is it?
On 22 September, Anthropic released Claude Opus 5.5. It is the first model of the new 5.5 family. According to Anthropic, it performs at the level of Fable 5.1 for most tasks, and costs 40% less to run than Opus 5. Smaller siblings Sonnet 5.5 and Haiku 5.5 will follow in a few weeks.
Shorter and clearer
This is the most noticeable change. Opus 5 received much criticism for producing answers that were lengthy and difficult to follow. Opus 5.5 is said to communicate more clearly and concisely than Opus 5 or Fable 5. Less text blocks. Quicker to the point.
Better at real work
Anthropic tested it on knowledge work, such as writing business reports. Opus 5.5 met the internal quality threshold in 16 out of 18 attempts. Fable 5.1 and Opus 5 failed every attempt. It also scores well independently: Artificial Analysis places it at the top of their Intelligence Index.
More accurate and faster
According to internal tests, the model is more truthful and less inclined to reason towards a desired outcome. It generates text over 30% faster than Opus 5, according to TechNews.
What we notice
We work with it daily. Blogs, research, fact-checking. The difference mainly lies in less need for adjustments. What used to take three rounds, now often gets right in one go.
To be fair: part of that probably also comes from the app. Chat and Cowork have been in one screen since this summer. Model and tool are hard to separate.
Small
It is the first Opus model where thinking cannot be turned off.
For sensitive questions about cybersecurity and biology, it reverts to the older Opus 4.8.
To follow
Sonnet 5.5 and Haiku 5.5, announced for the coming weeks.
More independent tests. Most figures currently come from Anthropic itself.
