# OpenAI SDKs add GPT-6.1 Sol and realtime translation support

Official SDK releases expose new model identifiers and additional agent and realtime API surfaces. Retrospective brief for 17:00–17:59 UTC on 29 September 2026.

## Python and Node SDKs add the GPT-6.1 Sol identifier

OpenAI Python 3.21.0 and Node 7.24.0 add the GPT-6.1 Sol model identifier. The release notes establish SDK support; account access and model availability still follow the service configuration. The Node release also adds agent credential and session options and Cyber access-program fields for Responses. 

[openai/openai-python v3.21.0 release notes](https://github.com/openai/openai-python/releases/tag/v3.21.0) · [openai/openai-node v7.24.0 release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)

## The Node SDK adds realtime translation connection support

OpenAI Node 7.24.0 adds client-secret creation for Realtime Translation and a dedicated Node translation WebSocket. It also adds opt-in incremental output snapshots for Responses WebSockets. These are explicit SDK API additions for applications that handle live translation or streaming model responses. 

[openai/openai-node v7.24.0 release notes](https://github.com/openai/openai-node/releases/tag/v7.24.0)

## Speculative decoding stops accepting draft tokens at end-of-generation

The llama.cpp b11259 development build stops accepting draft tokens once an end-of-generation token is reached. This is a correctness change in speculative decoding, where a draft model proposes tokens for verification. GitHub marks the build as a prerelease. 

[ggml-org/llama.cpp b11259 release notes](https://github.com/ggml-org/llama.cpp/releases/tag/b11259)