Blog Model Guides
DeepSeek V4 Pro and Flash Routing for Agents: A Decision Ledger
Give each routing decision a reason you can audit after the task finishes.
AIWave prices checked 2026-11-05 UTC against the live pricing response and active default-group routes. The old price and capacity examples on this URL were not current evidence. Check the live rate card before using a number in production.
| Exact model ID | Fresh input | Cache-hit input | Output |
|---|---|---|---|
deepseek-v4-flash | $0.30 | $0.006 | $1.20 |
deepseek-v4-pro | $1.32 | $0.044 | $3.96 |
Record the decision
For each agent step, store the task class, selected exact model ID, reason for selection, effective account group, request ID, retry count, and expected output check. After completion, add the actual usage, latency, result, and settled charge. Keep this ledger separate from public prompts or customer-identifying data.
Evaluate the policy
Use completed-task cost rather than a single request's listed token rate. A low-rate model that repeatedly fails or escalates can cost more than an initially stronger route. Compare the same task set under each policy, including retries and cache behavior, and keep the sample size visible.
Keep manual control
When a route or telemetry signal is missing, use a known, explicit model selection. A production policy needs an off switch, a bounded retry budget, and a way to trace every billed request to the chosen route. The current rate table below is only one input to that policy.
For an executable first request, use the quickstart. Recheck the effective account group and request settlement before treating an estimate as a bill.