devotd.
Recent developments, checked against their sources.
Gemini 3.8 TTS moves delivery instructions out of the transcript and into speech_metadata. Unary responses now default to WAV, so code that adds a header to raw PCM may break.
On September 30 NVIDIA announced general availability of the cuObject client and server libraries. Object payloads move to GPU or system memory over RDMA, but control requests stay on HTTPS/TCP, and storage servers must support the protocol too.
Google announced Gemini 4 Argon on September 30 with an output limit of 1 million tokens, up from 64,000. Access initially goes to trusted cyber defenders in its Fairwind Program; broader access was presented as a plan, without a general release date.
OpenAI’s September 29 announcement put GPT-6.1 Sol on the API and selected paid products. Cache reads are priced at $0.10 per million tokens, but cache writes cost $2.50; exceeding 272,000 input tokens raises rates for the entire request, not just the excess.