LLM Terminal: A Bloomberg-Inspired Market Intelligence Dashboard

3 Likes

I wanted to share a project I’ve been working on that pushes the boundaries of "traditional" dashboard design in SAS Visual Analytics. Inspired by the information-dense, dark-mode aesthetic of a Bloomberg Terminal, this report tracks the explosive growth of Large Language Models (LLMs) from 2017 to the latest 2026 frontier releases.

What makes this different?

Instead of wide-open white space, this dashboard embraces "Data Density." It’s designed for the power user who wants to see the whole market at a glance—from pricing wars to intelligence scaling. It uses a mix of native SAS VA objects and a custom Data-Driven Content (DDC) objects to create that high-frequency "trading floor" feel.

A Bloomberg-Inspired Market Intelligence Dashboard

The Tech Stack

The Data: Sourced dynamically from the OpenRouter REST API using PROC HTTP and processed in SAS Viya. It’s set up to refresh monthly so the "ticker" is always current with the latest releases (like GPT-5 or Llama 4).
The Tickers: Custom HTML/JS tickers embedded via DDC objects that receive real-time data from the SAS report.
The Theme: A modified "Midnight" report theme with custom monospace font styling (Courier New) to keep that retro-terminal vibe.

Visual Engineering & Precision

To capture the Bloomberg essence, the report utilizes a precision-layered architecture. I used a series of Precision Containers and Stack Containers to overlap visual elements, allowing for the "glow" effects and the dense integration of KPIs directly over charts. You'll see use of Display Rules that trigger neon-green and amber highlights for top-tier model performance. The background is pinned to a specific #111315 hex to ensure the neon data points truly pop against the terminal glass.

Decoding the Terminal (The Metrics)

If you aren’t an AI researcher, some of these acronyms might look like alphabet soup! Here’s what’s actually happening under the hood:

BMK (Benchmark Score): This is the "Intelligence Index." I’ve normalized various AI evaluations (like reasoning and math tests) into a 0–100 score. Think of it as the model's "IQ."
PRC (Price): Industry standard pricing for 1 Million Tokens. In this dashboard, I track "Input Price"—what it costs to send data to the model. It's a great way to see how "intelligence" is becoming a deflationary commodity.
SIZE (Parameters): Usually measured in Billions (B). This is the "physical" scale of the model's brain. More parameters usually mean better reasoning, but higher "Compute Load."
CONTEXT: The "Memory" of the model. A context of 128k means the model can "read" a whole book in one go; 1M+ means it can analyze an entire library of code.
MAX OUT (Max Completion): The limit on how much text the model can generate in a single response. Even if a model can "read" a million words, it might only be able to "write" 4,000 at a time!

Note on "Flavor" vs. "Fact"

In the spirit of replicating a true financial terminal, you'll notice some UI elements—like the specific telemetry numbers in the far bottom-left corner, the footer dashboard navigation buttons, or the "Analysis/Live Feed" toggle switch—are included primarily for aesthetic immersion. While they aren't currently linked to your CAS data, they help provide the "High-Frequency Trading" atmosphere that makes a terminal feel alive.

Enjoy!