Stop paying frontier prices for routine work.
Cortex routes every AI request to the cheapest model that passes your evals, enriched with your org's context before it runs.
+1Applications
Max budget$0.04
Applications
Max budget$0.01Applications
Max budget$0.60ApplicationsMax budget$1.20ApplicationsMax budget$0.02
Claude Sonnet 5AnthropicM01
Gemini 3 FlashGoogleM02Fable 5AnthropicM03
GPT–5.6 SolOpenAIM04
Grok 4 FastXAIM05
Cortex-triage-v3Cortex FoundryM06Every prompt is a buying decision.
Choose among eligible models using your evals, budget, and latency target.
Triage this support escalation, with prior resolutions
Route candidates
- Eligible
General model A
Passes evals · Higher cost
- SelectedPrivate support model
Passes evals · Lowest eligible cost
- Excluded
General model B
Exceeds latency target
Match the model to the task
The cheapest path that passes your evals, within your budget and latency target.
Enriched before it runs
Retrieve relevant, permission-scoped context before the model runs.
Apply your routing rules
Check cost, latency, and access requirements before selecting a supported route.
Sparse prompt in. Your whole org behind it.
At run time, Cortex searches your company context graph and memory stores that Atlas builds. The task, the artifacts, and the history attach before the model runs. Ten words in, a grounded answer out, at a fraction of the cost.
“Triage this support escalation”
Priya NairL2 support
SEC-1204Exposed key rotationPR #4821rotate-keys +312 -88#support-esc14 messages · 22m agoRunbook / Key rotation v3Last update Jul 2
3 similar resolutions96% accepted
Acme CorpEnterpriseEU regionPrior fix / cortex-triage-v3Accepted Jun 12Other routers read the prompt. Cortex knows the work.
Cortex picks the model using signals no prompt-based router has: the task being done, the person's role and permissions, and which model produced accepted work on that task before.
Without CortexRaw prompt
GPT-5.6 Sol41 tokens$0.42 / call|
With CortexEnriched prompt
GPT-5.6 Sol1,286 tokens$0.42 / callSEC-1204PR #4821#support-esc|
Cortex Router · tau2-bench
247/276 tasks passed (89.5%) at $0.022 per task for the Cortex OSS policy.
Cortex Router vs Sol
500 SWE-bench Verified tasks: 75.2% vs 79.2% quality credit, at 27% lower model cost per task.
Cortex Router vs OpenRouter
400-prompt Claude-tier routing comparison. Same study as Not Diamond.
Cortex Router vs Not Diamond
400 prompts: 91% of fixed Opus quality at 40% lower normalized cost per prompt.
See Smart Routing & Enrichment on your own data.
We'll scope it as part of Cortex Foundry on your stack in days.