Choose the right model for the job.
Describe your task and how private it has to be. We combine independent quality scores and task-completion time with live provider routes, integer pricing, privacy posture, and measured route latency.
Give your coding agent the model picker.
Codex, Claude Code, Hermes, Cursor, and other agents can use the
trustedrouter-model-advisor skill to choose models with AI IQ task speed,
live cost, privacy, context, provider health, and prompt-cache context.
Read https://raw.githubusercontent.com/Lore-Hex/LLM-advisor/main/SKILL.md
Choose a TrustedRouter model for this task.
Estimate cost and speed before making billable calls.Three axes, one honest tradeoff.
Smart uses independent AI IQ scores. Cheap uses integer microdollar prices for the least expensive route that clears your privacy floor. Fast uses AI IQ's completion-time model: initial response time plus the time required to finish a standard 1,000-token answer. TrustedRouter's measured route TTFT stays visible as separate delivery context.
Pick a privacy floor — Open, Zero-retention (ZDR), or a hardware Trusted Execution Environment (TEE) with end-to-end encryption — and the board only keeps models with an exact provider route that clears it. The TrustedRouter gateway hop is attested at trust.trustedrouter.com.
base_url="https://api.trustedrouter.com/v1"
# let the router pick, per request:
model="trustedrouter/auto" # general rollover, no upstream privacy floor
model="trustedrouter/fast" # lowest measured TTFT among healthy candidates
model="trustedrouter/cheap" # lowest-cost paid candidate pool
model="trustedrouter/eu" # EU-focused provider pool
model="trustedrouter/zdr" # zero-retention or stronger routes
model="trustedrouter/confidential" # provider confidential compute + E2EE
model="trustedrouter/synth" # billable panel + judge + synthesizer
model="trustedrouter/advisor" # worker with optional billable advisorsCompare the frontier you can actually route.
The chooser loads current models and provider endpoints from the same deployed catalog as
/v1/models. TrustedRouter's auto, fast, cheap,
confidential, and orchestration routes expose those choices through one
OpenAI-shaped endpoint. Gateway attestation verifies the router hop; endpoint rows state the
downstream provider posture separately.
Questions
How do you decide which model fits?
Tell us the task and we estimate the intelligence it needs (simple to frontier), the latency you can tolerate (real-time to overnight), and a privacy floor. We keep only models with an exact provider route that clears the selected floor, then rank them by the smart/cheap/fast tradeoff.
What do the privacy tiers mean?
The TrustedRouter gateway hop is attested on every request. Open permits any upstream posture. Zero-retention (ZDR) requires a provider endpoint whose verified policy or contract retains nothing. TEE requires provider confidential compute plus provider-side end-to-end encryption.
Which models are fastest?
The picker uses recent measured median throughput and time to first token for the exact provider routes that satisfy your privacy selection. Unmeasured routes are labeled instead of receiving an invented speed.
Do I have to pick one model?
No. trustedrouter/auto provides general rollover without an implicit upstream privacy floor, trustedrouter/cheap chooses low-cost paid candidates, trustedrouter/zdr enforces zero retention, and trustedrouter/e2e enforces confidential compute plus provider-side encryption.