Cortega Model Recommender for Enterprises

Industry & application

Pick what's closest to your situation to prefill Workloads & scale below, or skip straight there.

Industry optional

Application

Workloads & scale

Prefilled from the application above — adjust anytime. Deselect all to enter usage yourself.

Workloads select all that apply — none is allowed

Team size

People / seats this estimate covers

Estimated usage

Pre-filled from workloads and team size as monthly totals. Enter measured input and output tokens per month — you do not need to divide by request count.

Input tokens / month Output tokens / month

In:out ratio—

Base models optional

Enter model names only (not sellers or providers). Add every model you already run. Ranking always uses the weight sliders below.

Add one or more base models

Weights

These sliders start from the workloads you selected. Move them to change how models are ranked.

Quality Cost Coding Reasoning Chat Speed Tools

Only open-weight models Show my base models at the top of the list Max monthly budget ($) Show cost mode

Locally Hosted shows the memory a model needs on hardware you already have or are deciding to buy — not a cost estimate. It uses vLLM on Intel Arc, NVIDIA, and AMD GPUs — faster than Ollama's defaults, which most people running local models haven't tried. Apple Silicon has no GPU-accelerated vLLM path yet, so that one runs on llama.cpp / MLX with Metal acceleration instead — still well ahead of Ollama's defaults, just a different engine. There's no one-click installer for local hardware yet (today's self-hosted-LLM wizard provisions AWS + NVIDIA only).

Recommendations

Ranked from the curated catalog for the inputs above.

This is a one-time snapshot. Cortega Model Manager continuously watches your actual traffic and suggests model upgrades or replacements automatically, instead of you re-running this by hand.