In the latest independent global ranking for spreadsheet performance, Decide Agent 2.0 scored 33.81% overall.
Ahead of:
GPT-5.2 — OpenAI
Gemini 3.1 Pro — Google
GLM-5.0 — Zhipu AI
DeepSeek V3.2 — DeepSeek
Kimi K2.5 — Moonshot AI
Qwen3.5-397B-A17B — Alibaba
MiniMax M2.5 — MiniMax
SpreadsheetBench V2 measures agents across financial modeling, template completion, spreadsheet debugging and visualization, making this a strong result on end-to-end business spreadsheet workflows.
We achieved this with DAX-1 and months of improving the product by listening closely to our users and building with them.
Agent 2.0 is now live in Decide on the web, in Excel, Google Sheets and via the API.
Decide for Excel is powered by DAX-1, our specialist model for spreadsheet operations with 91.7% accuracy, 5× faster execution, and 97% lower operating cost.
It works directly inside Microsoft Excel, so you can analyze data, edit spreadsheets, and automate workflows without leaving your workbook.
https://t.co/m6TQAG2cCw
Our first case study is about a startup that replaced Gemini with Decide for document extraction.
PharmacyOS was using Gemini to extract structured data from handwritten pharmaceutical orders, images, and receipts, but ran into issues with accuracy, consistency, and cost at scale.
They switched to the Decide Agent API, resulting in more reliable structured extraction with economics that make more sense for a production workflow.
Our edge is our ability to deploy specialized AI solutions that can perform specific workflows better and more efficiently than general-purpose frontier models.
We built the Decide API so businesses can plug these specialized capabilities directly into their products and production workflows.
DAX-1 is now available as open weights for non-commercial research.
It reaches 91.7% accuracy on our 350-task spreadsheet-editing benchmark, compared with Fable 5’s 58.3%, at 97% lower benchmark cost and over 5× faster median response time.
The open-weight release makes DAX-1 available for researchers and builders to inspect, run, and reproduce.
https://t.co/YTDaYE55QU
PharmacyOS replaced Gemini with the Decide Agent API for document extraction.
The team needed to extract structured data from handwritten pharmaceutical orders, images, and receipts, but ran into issues with accuracy, consistency, and cost at scale.
With Decide, PharmacyOS now gets more reliable structured extraction with economics better suited for a production workflow.
The Decide API gives businesses specialized AI capabilities they can integrate directly into products and production workflows, with better control over reliability, performance, and cost.
Read the full case study:
https://t.co/WjWPqC1eEH
We are announcing our most ambitious move today with the release of DAX-1, our flagship model, post-trained specifically for spreadsheet editing.
DAX-1 is 97% cheaper to run than Anthropic’s Fable 5 and achieved 91.7% accuracy compared to Fable 5’s 58.3% on our benchmark. DAX-1 is also 5× faster.
This marks a shift in how we build Decide. Rather than relying solely on general-purpose frontier models, we will train specialized models where it makes sense for our operations.
This gives us more control over accuracy, speed, and cost, and allows us to build better products for the businesses and professionals who rely on Decide.
DAX-1 is the first step in this direction
If you’re Nigerian, there’s a very high chance you’ve experienced or witnessed some form of insecurity.
I got tired of just talking about it. So I’ve been building.
This is ResQ.
https://t.co/VBdJ8TCT2v
@DaveyHert@akinkunmi@victorfatanmi
Happy to announce that we have completed the training of our flagship model, Dax-1, which is entirely focused on deterministic spreadsheet editing.
This marks a major shift in how @trydecideAI operates and is, by far, the greatest engineering feat we have undertaken as a team of three.
We achieved remarkable performance, and we will share further details about the training process and results through technical reports and benchmarks when we go live later this month.
Skills are now available in Decide for Excel.
Skills let you save repeatable workflows, so the agent performs the same task exactly the way you want every time.
For example, create a Monthly Sales Report skill that cleans your data, builds pivot tables, and formats the report with a single command.
Simply type / to run a saved skill.
Decide for Excel will be generally available soon.
Decide for Excel and Google Sheets are now available to our first group of users.
We’ll be rolling out access in batches.
Over the next few days, we’ll be showing what it can actually do inside your spreadsheets.