Change scope
How many files and lines, and whether it crosses modules
Please wait while the page loads.
Stage-based model routing
Code search and dependency reading run on deepseek-flash; the refactors and debugging that decide quality go to deepseek-v4-pro. You never switch models by hand: DeepCodex decides per stage and records the reason, spend and cost of every choice.
Decision signals
Not random and not by price alone. Every choice shows its reasoning in the request detail.
How many files and lines, and whether it crosses modules
Looking up a fact, or comparing several designs
Whether it touches payments, permissions, migrations or other sensitive paths
Whether this task and this period can still afford the deep model
In short:light signals → deepseek-flash;heavy signals and budget allows → deepseek-v4-pro;When the budget is short, results are never silently downgraded: the run stops and returns an explicit reason_code (for example quota_exceeded).
Savings estimator
Enter tokens per stage; three baselines side by side: deep throughout, fast throughout, mixed.
CN¥0.06 (28%) less than deep throughout
equiv. tokens = fast tokens × 1× + deep tokens × 4×
Estimated at peak prices without the cache discount; the request detail is authoritative. deepseek-flash / deepseek-v4-pro
Three working modes
The mode is picked by the platform per stage difficulty; the request detail records the model and spend of each stage.
Search, summaries, formatting and small edits
Architecture, hard debugging and large refactors
Analysis first, core reasoning, then tidy-up
Pipeline
The fast model does the frequent tidying, the deep model the quality-defining core, then the fast model wraps up.
Read the task, extract the file scope, list known constraints
Compare designs, reason across files, generate the hard code
Format checks, change summary, structured result
equiv. tokens = fast tokens × 1× + deep tokens × 4×. However many times the model switches, the bill has one unit.
Routing policy
Each key can restrict the model allowlist, concurrency, RPM and period cap. Automatic routing works only within the allowed models and quota.
Automatic routing from the first request; every model switch is traceable in the request and task detail.