OpenAI o1-preview API deprecated since 2025, reasoning parameters now in GPT-5.6 series

OpenAI removed the o1-preview model from its API on July 28, 2025. Developers previously using the preview for multi-step reasoning tasks were directed to o1 and later o3 series as replacements.

The full o1 model launched in December 2024 with the reasoning_effort parameter in low, medium and high settings. This allowed control over the number of internal reasoning tokens, affecting latency and cost.

Model shutdowns and current options

On October 23, 2026 OpenAI shut down remaining o-series models including o1 and o3-mini. The recommended replacement is gpt-5.6-sol with pro reasoning mode. Current GPT-5.6 and GPT-6 families support expanded values for reasoning.effort: none, minimal, low, medium, high, xhigh and max.

Reasoning tokens are billed as output tokens. The Responses API provides better performance than Chat Completions for these models. Russian developers access compatible endpoints through local aggregators that accept RUB payments because direct OpenAI API access remains unavailable since February 2024.

Practical migration steps for developers

Teams should test replacements against their existing prompts to measure changes in reasoning token usage and output quality. Higher effort settings increase both quality and cost, so monitoring token consumption remains essential after migration.

Documentation on developers.openai.com lists the supported parameters and model-specific defaults. No Russia-specific expansions of o1-preview occurred before its removal.