Supported AI Models

Deep reasoning, complex coding, long multi-step tasks.

Maximum accuracy for complex reasoning and research.

Slightly weaker GPT-5.2, cheaper, still strong at reasoning.

Early GPT-5 release. Mostly superseded by newer 5.x models.

Cheaper GPT-5 for structured tasks and reasoning at scale.

Ultra-cheap GPT-5 for classification, extraction, and routing.

Legacy high-compute GPT-5. Replaced by GPT-5.2 Pro.

GPT-5.2 tuned for ChatGPT conversations.

GPT-5.1 conversational model used in ChatGPT.

Early GPT-5 chat model. Mostly superseded.

Long-context text and image processing. No deep reasoning.

Cheaper GPT-4.1 for large-scale document processing.

Lowest-cost long-context model for extraction and tagging.

General coding model optimized for IDE-style workflows.

Faster, cheaper coding model for refactors and linting.

High-accuracy Codex for large codebases and migrations.

Older Codex model. Mostly replaced by 5.1 Codex.

Research reasoning for large document analysis.

Multi-step reasoning for analysis, research, and planning.

Lightweight reasoning with good cost-accuracy balance.

Core reasoning model for math, logic, and structured thinking.

Cheaper reasoning model for constrained logic tasks.

Higher-accuracy o3 for complex reasoning workloads.

Early reasoning model for basic logic and step-by-step tasks.

Legacy reasoning model. Weaker than o3-series.

Early lightweight reasoning model. Kept for reference.

Experimental reasoning model. Do not use.

Real-time chat and multimodal apps with low latency.

Cheaper GPT-4o for high-volume chat and UI assistants.

Large-context GPT-4. Slower and pricier than newer models.

Original GPT-4. Kept for legacy compatibility.

Experimental GPT-4 variant. No longer recommended.

Low-cost model for simple chat, classification, and extraction.