Decide Agent 2.0 tops SpreadsheetBench V2 at 33.81%, ahead of GPT-5.2 and Gemini 3.1 Pro
saheedniyi_02 · x · 2026-09-07
- Decide Agent 2.0 scored 33.81% on SpreadsheetBench V2, ahead of GPT-5.2, Gemini 3.1 Pro, GLM-5.0, DeepSeek V3.2, Kimi K2.5, Qwen3.5-397B-A17B and MiniMax M2.5.
- The benchmark covers financial modeling, template completion, spreadsheet debugging and visualization — end-to-end business spreadsheet workflows.
- Built on the in-house DAX-1 model and months of user-feedback iteration; live on web, Excel, Google Sheets and API.
More from Apps
- Indie Dev's AI Product Vinu Hits Email Sign-up Goal Ahead of Launch — pramodk73 · 2026-09-07
- Tsinghua's OpenMAIC Turns Any PDF into an AI Classroom, 32k GitHub Stars — 量子位 · 2026-09-07
- Pavo Launches Instant Video Mode: 10-Second Clips, No Signup, Free Flash Model — 机器之心 · 2026-09-07
- One Swede built Marginalia, an ad-free Google alternative with its own index — thisguyknowsai · 2026-09-07
- roadmap.sh: open-source roadmaps map the learning order for every tech job — bigaiguy · 2026-09-07
- AgentInflux early version recommends AI agents that fit your business workflow — AgitatedBed8262 · 2026-09-07