GPT 5 Codex: A Specialized Model for CLI Use
OpenAI released gpt-5-codex on September 23, 2025. This model is optimized for the Codex CLI and accessed via the v1/responses API with text and image inputs.
Release Announcement and Scope
OpenAI announced the release of gpt-5-codex on September 23, 2025. This date marks the official launch of the model as recorded in the provider's changelog. The model is identified as a special-purpose tool designed specifically for use with the Codex CLI. It is part of the GPT family and is accessed through the v1/responses API endpoint.
- Model Name: gpt-5-codex
- Provider: OpenAI
- API Endpoint: v1/responses
- Primary Use Case: Codex CLI
Confirmed Capabilities and Evidence Limits
The source facts confirm that gpt-5-codex accepts both text and image inputs. Its output modality is limited to text. The documentation does not specify a context window size or maximum output token limit. Additionally, open weights are not confirmed for this model. There is no information provided regarding pricing structures or local installation requirements. The release notes do not describe internal architecture or specific algorithmic improvements over previous iterations.
- Input Modalities: Text, Image
- Output Modality: Text
- Context Window: Not specified
- Max Output Tokens: Not specified
- Open Weights: Not confirmed
Proposed Evaluation Procedure
To assess the practical utility of gpt-5-codex, engineers should conduct controlled tests within the Codex CLI environment. First, define a set of representative coding tasks that align with the intended workflow. Second, measure the model's response accuracy and relevance against these tasks. Third, compare the results with baseline expectations or previous tools used in similar contexts. Finally, validate whether the model’s behavior meets specific project requirements. This process helps determine if the specialized optimization provides tangible value for command-line interactions without assuming unverified benefits.
- Test: Execute standard coding prompts via Codex CLI
- Measure: Record success rates and error types
- Compare: Contrast outputs with existing workflows
- Validate: Confirm alignment with team standards
Artificial Analysis benchmark review
Benchmark review updated on . These scores apply to the linked artifact and recorded methodology; they do not establish the performance of a different runtime or quantized model.
| Measurement | Recorded result | Measured |
|---|---|---|
| Artificial Analysis Intelligence Index (v4.1) ↗ | 37 index_points | 2026-08-09 |
| Artificial Analysis Intelligence Index (v4.2) ↗ | 29.2 index_points | 2026-09-05 |
| Artificial Analysis Intelligence Index (v4.3) ↗ | 24.9 index_points | 2026-09-09 |
Review the full benchmark provenance and source methodology →
Primary sources
Frequently asked questions
What is the primary function of gpt-5-codex?
gpt-5-codex is a special-purpose model built and optimized for use with the Codex CLI. It is designed to handle text and image inputs to produce text outputs within that specific command-line interface.
Can I install gpt-5-codex locally?
The source facts state that open weights are not confirmed. Therefore, there is no evidence provided to establish whether local installation is possible or impossible. Access is documented via the v1/responses API.