ModelGPT-5.2-Codex

GPT-5.2-Codex

A code-specialized large language model version released by OpenAI, optimized for programming and code generation tasks and commonly used in agent programming scenarios.

Timeline (last 90 days)

Oct 24

LangChain's Deep Agents, on GPT-5.2-Codex, raised its Terminal-Bench 2.0 score from 52.8% to 66.5% solely by restructuring the harness.

Unverified50%

All Facts (1)

Unverified

LangChain's Deep Agents, on GPT-5.2-Codex, raised its Terminal-Bench 2.0 score from 52.8% to 66.5% solely by restructuring the harness.

50%

Source Articles

Same Model, 22-Point Score Gap: Agent Harness Is the Key VariableOct 24