Decido is listed as Open Source. Check https://github.com/decido-ai/decido for the current plan limits, as vendors change pricing frequently.

Decido

Q: What is Decido?

Decido is a code AI tool listed on Use AI Tools. Decido returns calibrated yes/no, pick-one and rating decisions about a document on a laptop CPU. You hand it a state — text, JSON or a list — and typed questions, and it returns a probability for every answer in the same wire format as TypeSafe's Jev System One API. The model is a LoRA fine-tune of Gemma 4 26B-A4B, about 4B active parameters, shipped as an 8-bit GGUF for llama.cpp; install.sh builds llama.cpp, downloads the weights and runs an example. Question types cover yes or no, a choice of 2 to 255 options, and ordered score levels.

Q: What is Decido best for?

Decido is commonly used for Turning a document into typed decisions, Running decision models on CPU only, Keeping the Jev wire format you already use. It supports English.

Code⚡ Advanced✅ Global access

Turning a document into typed decisionsRunning decision models on CPU onlyKeeping the Jev wire format you already use

Open SourceEnglish

Decido returns calibrated yes/no, pick-one and rating decisions about a document on a laptop CPU. You hand it a state — text, JSON or a list — and typed questions, and it returns a probability for every answer in the same wire format as TypeSafe's Jev System One API. The model is a LoRA fine-tune of Gemma 4 26B-A4B, about 4B active parameters, shipped as an 8-bit GGUF for llama.cpp; install.sh builds llama.cpp, downloads the weights and runs an example. Question types cover yes or no, a choice of 2 to 255 options, and ordered score levels.

Visit Website Compare

Community Rating

Be the first to rate Decido. Your feedback helps other users decide.

🎯 Real-World Use Cases

Ask a yes or no

Send a support ticket as the state and ask whether the customer wants a refund; get a probability back.

Route to a team

Use a choice question with criteria per option to get probabilities, an argmax and a confidence.

Run it locally

Build llama.cpp through install.sh and serve the 8-bit GGUF on CPU with no hosted API.

⚖️ Honest Review

✅ Pros

+Same input always gives the same answer on a machine
+Runs on CPU with a published GGUF
+Trained on public data and an Apache-2.0 teacher model

⚠️ Cons

−About 30 GB of RAM and a few seconds per call
−Numbers quoted are the project's own runs, not an official benchmark

📚

Guides and comparisons about code tools

CodeOct 26, 2026