RUST · AI BENCHMARKING

decision-index — Reproducible Latency & Accuracy Benchmark for AI Decision Engines

decision-index is a standardized benchmark suite that compares sub-microsecond deterministic Rust decision matrices against local and cloud LLM inference. It measures p50/p99 latency, memory footprint, cost per million decisions, and classification accuracy across real-world routing and triage workloads.

$ cargo run --release

GitHub ↗

Zev SIMD Latency 5.8 µs

LLM Router Baseline 850.0 ms

Determinism 100.0% Exact

Cloud Token Cost $0.00 / 1M ops

Interactive Decision Router Latency & Cost Calculator Routing Intent Query / Classification Vector Comparison Baseline

Deterministic Execution Output & Telemetry STATUS: READY // 0 ERRORS

Benchmark Specification

{
  "suite": "decision-index-v1",
  "workloads": 10000,
  "results": {
    "zev_simd": { "p50_us": 5.8, "p99_us": 7.1, "accuracy": 0.994, "cost_usd": 0.0 },
    "gpt4o_router": { "p50_us": 850000, "p99_us": 1420000, "accuracy": 0.991, "cost_usd": 30.0 }
  }
}

Production Engineering Specifications

Repository:    https://github.com/bhubbard/decision-index
Live Studio:   https://code.brandonhubbard.com/decision-index/
Portal:        https://code.brandonhubbard.com/
Install:       git clone https://github.com/bhubbard/decision-index
Design System: Brandon Hubbard Swiss Modernist (#0A0A0C / 1px Exposed Structural Grid)