LLM Routing, Cost & Reliability — Developer Research
CarbonLayer is researching how developers manage LLM inference costs, latency, provider reliability, and model selection. This anonymous, voluntary survey takes 3–4 minutes. No name or email required. Please don't include confidential information. If you select 'Not currently, but planning to' or 'No,' you may scroll to the bottom and submit the survey without answering the remaining questions.
Are you currently building or operating an application that uses LLM APIs?
Yes, in production
Yes, in development or testing
Not currently, but planning to
No
Other
Think about the last three months. Describe a specific LLM cost, latency, provider reliability, or model-selection problem you encountered. What happened, and how did you handle it?
How does your application currently choose which model or provider handles a request?
One fixed model or provider
Routing or fallback logic written in application code
Configurable gateway or routing platform
Manual selection based on workload
Another approach
Not sure / Not applicable
Which controls have you actually configured or implemented?
Cost or token budgets
Latency thresholds or timeouts
Provider fallback or circuit breakers
Provider or model restrictions
Quality-based model selection or escalation
Per-request cost or routing telemetry
Which tools, services, or custom implementations are you using for LLM routing, cost monitoring, or budget enforcement?
Is there anything you still find difficult to control or configure in your current setup? If so, what workaround are you using?
Which single issue, if any, would be most valuable for you to improve?
*
Reducing unpredictable inference costs
Meeting response-time requirements
Handling provider failures or rate limits
Enforcing provider or model restrictions
Maintaining acceptable response quality
Understanding or auditing routing decisions
Something else
No significant issue currently
Submit Survey
Should be Empty: