AI Agent StorevLLM Semantic Router · Download documented
d3 · local decision models
Route support requests, prioritize cases or check an image against a rule on your own machine. Supply the possible answers; d3 scores them instead of writing a reply.
First reviewed 2026-10-27 · Checked 2026-11-03. The launch date has not been confirmed.
We still need comparable independent tests. You can compare the confirmed features and limits below, but we cannot yet give this version an Agent Fit or Practical Value rating. Try it on your hardware and tasks before relying on it.
What you can use today
You download the model files and supply the hardware and software to run them. Capacity is the amount of information a request can contain; settings and available memory can limit what fits.
Exact Ollama package ID: d3:2b / d3:4b / d3:9b / d3:27b. The four packages contain 2B, 4B, 9B and 27B variants. The plain d3 name selects 4B. Ollama supports text, structured data and still images; video and Box/Track decisions are unavailable in these packages.
Requires Ollama 0.40.9 or later. The page recommends Apple Silicon systems with 8GB, 16GB, 32GB and 64GB respectively, based on its default packages at 4,096 context. Other systems use a different format and may need different memory. Apache 2.0.
The package instructions specify 4,096 context tokens despite the larger 256K catalog badge. Do not assume the badge is the usable default. Publisher checkpoint capabilities and scores do not establish equivalent behavior in these compressed packages.
No per-token local API bill is established; you supply hardware and electricity. Check probabilities on real examples before setting thresholds for automatic actions.
Provider release and specifications ↗
Check availability and plans ↗
More setup and licensing details ↗
More setup and licensing details ↗
More setup and licensing details ↗
Test it before switching
Give each model the same examples, tools and definition of success. Record correct outcomes, response time, retries and total cost. Include the time and hardware needed to run it.
Compare caching and long-context costs ↗