Claude Sonnet 5.5 makes everyday work faster and less costly

Anthropic released Claude Sonnet 5.5 on September 28. The model is designed to help with everyday tasks such as writing, preparing documents, and fixing code at a faster pace.

Finish the same work in fewer steps

Anthropic says it is more than 30% faster than Sonnet 5 and can cost up to 30% less per completed task for typical work. The per-token price has not fallen; the company attributes the difference to fewer steps and fewer tokens. It also highlights documents, slides, and spreadsheets alongside coding.

The model is rolling out across Claude products and third-party services. It is also available in GitHub Copilot, although access there is being expanded gradually.

More room to build, test, and revise

A team can make a screen, try it, then revise it. The same applies to a document after seeing how readers use it. Faster and cheaper work makes those cycles easier to repeat.

The comparisons come from Anthropic’s tests and do not promise the same result for every task. Still, models capable of complex work are becoming practical for smaller daily improvements.

anthropic.com ↗ ↗

More chances to try and revise.

What stands out in Sonnet 5.5 today is how much lighter it may make the full task, beyond the benchmark scores.

Make a screen. Try it. Then change the part that does not work.

That cycle is becoming less dependent on a separate development project. Runway now lets editors bring generated footage straight back into their editing timeline. Proaction uses Codex to reshape product demos for customers.

The wait between asking someone else to make a change and seeing the result is getting shorter.

More chances to look and revise mean the first draft does not need to carry the burden of being final. See how people respond, then build again. That loop becomes part of everyday work.

Quality will increasingly come from the number of times a team can test and refine an idea.

Claude Code / Anthropic

No new article was verified in this category today.

OpenAI / ChatGPT

Medium impactMedium signal
Codex helps tailor product demos for each customer
North American fleet software company Proaction says it uses Codex to tailor product demos to customers. Its sales team can move forward with changes that previously required help from engineers.
openai.com ↗ ↗

AI Agents / Coding

Medium impactMedium signal
Mistral brings questions and tasks into one Vibe workspace
Mistral announced on September 22 that Vibe Chat and Work are merging. Users can ask questions and start tasks from the same place, with a shared history and reusable instructions. Migration is gradual.
docs.mistral.ai ↗ ↗

AI INSIGHTS / KOHEI TORIGOE

Kohei Torigoe

AI Product Architect

Over 25 years in UI/UX design, product development and business development. Founded and led a company with a team of around 100 people.

Currently planning, designing and developing new products using AI.

Through AI Signal, I share discoveries from developing new products and the possibilities AI opens up for the future, while trying the latest AI technologies firsthand.

Profile, work and contact ↗

Video Generation AI

Medium impactMedium signal
Runway brings generated footage into DaVinci Resolve
Runway released a plugin for DaVinci Resolve Studio. Editors can generate or restyle footage and bring the result back to the timeline. It requires a paid Runway plan and Resolve Studio 19 or later.
runway.com ↗ ↗

AI Music Generation

No new article was verified in this category today.

Foundation Models / Research

No new article was verified in this category today.

AI Automation / Business Use

Medium impactMedium signal
Ringg routes AI customer calls across models and languages
Ringg published a case study on using several OpenAI models for calls and chat. It tests models by language and region and says some inquiries are handled without a human agent. The outcomes are company-reported.
openai.com ↗ ↗

AI Insights