5.0.0

5.0.0

Breaking changes and migration

• Image variations: Remove Request-ImageVariation. The /images/variations endpoint seems to be deprecated by OpenAI.
• Default models: Use gpt-6-luna for Request-ChatCompletion, Enter-ChatGPT, Request-Response, and Request-ResponseCompaction. Use gpt-transcribe for Request-AudioTranscription and Realtime input audio transcription.
• Azure OpenAI: Remove the Azure-specific API mode and ApiType, ApiVersion, and AuthType parameters/context properties. Remove these parameters from scripts, including -ApiType OpenAI. For Azure v1, set ApiBase to https://<resource>.openai.azure.com/openai/v1/, pass an Azure API key or Entra ID token as ApiKey (Bearer authentication), and use the deployment name as Model. Legacy deployment URLs and api-key authentication are no longer used. See the migration guide.
• Video API (Sora): Remove New-Video, New-VideoRemix, Get-Video, Get-VideoContent, Remove-Video, and Wait-Video. There API is no longer available from OpenAI.
• Local utilities: Remove ConvertTo-Token, ConvertFrom-Token, and Get-CosineSimilarity, including the bundled tokenizer library. Use another library for tokenization or vector similarity.
• API keys: Explicit -ApiKey $null now uses an empty key. Omit -ApiKey to use the configured key.

Added and improved

• Add Request-ContentProvenanceCheck for image and audio provenance checks (C2PA and SynthID).
• Add completions for gpt-6-sol, gpt-6-luna, and GPT Image 2.5 models. Accept additional image quality and size values supported by the server.
• Add Request-Response -RemoteMCPTunnelId, -PromptCacheComparisonResponseId, and -PromptCachePrewarm. Deprecate -ConnectorId for models released after September 1, 2026.
• Share the HTTP transport across regular and streaming requests. TimeoutSec now covers each attempt's response read.

Fixed

• Fix API key precedence, response compaction prompt cache options, and structured output parsing of final-answer content.