Skip to content

Inference tiers tuned to the work.

Lite, Pro, and Max are operating profiles, not a list of model names. Lite handles real-time conversation, Pro handles code and precision work, and Max handles long-context reasoning. The endpoint stays the same.

The workload changes.
The endpoint does not.

Lite, Pro, and Max are operating profiles—not a list of model names. Each one is tuned around the constraint that matters most to the work.

01lite-64k

Speed & volume

Lite

Real-Time FastLow-latency inference for customer-facing chat, high-volume classification, and real-time AI agents.Explore Lite
64Ktoken context window
Real workloadResolve a live customer request.Latency first
Input
Support history + current message
Route
lite-64k
Output
Low-latency streamed answer
02pro-256kMost popular

Code & precision

Pro

Code & PrecisionStructured generation accuracy, multi-file coherence, and instruction fidelity for precision-sensitive workloads.Explore Pro
256Ktoken context window
Real workloadRefactor a multi-file service.Precision first
Input
Repository context + instructions
Route
pro-256k
Output
Structured patch + tool calls
03max-1mFlagship

Reasoning & context

Max

Flagship ReasoningLong-context reasoning, multi-document synthesis, and multi-step agent orchestration at scale.Explore Max
1Mtoken context window
Real workloadAnalyze an entire codebase.Context first
Input
Full repository + extended history
Route
max-1m
Output
Cross-system synthesis and plan