DVAR MODELS / OPEN-WEIGHT LIBRARY
Choose the smallest model that can do the job.
A practical catalog for Indian enterprise teams. Start with capability and economics, then validate the model on your own workload before you commit.
A model is only “best” after it wins on your data, latency, and budget.
DVAR MODEL LIBRARY / WORKING CATALOG
Open models, with the details that matter in a buying conversation.
Our internal catalog organizes open-weight model families by capability, context, license, and deployment fit. We keep it current as model cards and production checks change.
Model names, context windows, licenses, and deployment availability change. This page is Dvar’s working catalog and decision aid, not a promise that every model is available in every deployment mode. Dvar benchmarks the selected model on your workload before recommending production use.
HOW TO CHOOSE
Quality is not one number.
Use the workload.
Choose Eco when volume matters first
Use smaller models for classification, extraction, routing, and high-volume employee work where a 10% quality gain may not justify a 4× token bill.
Choose Balanced when the task is mixed
Use a capable 8B–32B model for everyday assistants, structured outputs, and production workloads with a real latency budget.
Choose Pro when the decision is expensive
Use larger reasoning or agentic models for complex research, code, and workflow design, then decide whether a tuned smaller model can take over the repeatable core.
MODEL EVALUATION
Bring us your benchmark.
We will bring the shortlist.
Share 20–50 representative examples and your latency and budget constraints. We will return a decision-ready comparison, not a leaderboard tour.
Contact Us