Data & AIAI Models
Model Catalog & Routing Engine
Registered models, benchmarks and smart routing policies powering the AI cost optimization engine.
Registered models
20.0%
6
Across 4 providers
Routing policies
17
17 active workflows
Avg quality score
1.4%
91
Weighted by traffic
Savings via routing
6.6%
32%
vs all-flagship baseline
Model catalog
Benchmarks, cost & traffic share
| Model | Tier | $ / 1M tok | p95 (ms) | Quality | Traffic % | Status |
|---|---|---|---|---|---|---|
GPT-4o OpenAI | Flagship | $5.00 | 820 | 96 | 41% | Live |
Claude 3.5 Sonnet Anthropic | Flagship | $3.00 | 720 | 95 | 24% | Live |
Claude 3 Haiku Anthropic | Fast | $0.25 | 240 | 84 | 18% | Live |
Gemini 1.5 Pro Google | Flagship | $3.50 | 1040 | 92 | 9% | Live |
Dust-RAIP-7B Dust Internal | Specialist | $0.08 | 180 | 88 | 6% | Live |
Llama 3.1 70B Dust Internal | Open weight | $0.60 | 480 | 89 | 2% | Review |
Quality vs cost
Benchmark score per dollar
Smart routing policies
Workflow → model assignment with fallback chain
Tier-1 Support Triage
low-latency · cost-sensitive
Claude 3 Haiku
fallback → GPT-4o mini
Contract Intelligence Review
long-context · reasoning
Claude 3.5 Sonnet
fallback → GPT-4o
Invoice Reconciliation Pipeline
fine-tuned · domain
Dust-RAIP-7B
fallback → Claude 3 Haiku
Clinical Note Summarization
PHI policy · accuracy
Claude 3.5 Sonnet
fallback → GPT-4o
Pharma Trial Data Extraction
structured extraction
GPT-4o
fallback → Gemini 1.5 Pro