Friday tech news: three updates for builders
Friday tech news: three updates for builders 1. OpenAI's Decisions API is generally available (Oct 22) What changed: The Decisions API is GA, built on GPT-6 Luna with a dedicated endpoint for three answer types: predicate (a probability), choice, and score. Pricing is $0.10 per million input tokens, $0.01 per million cached input tokens, and no output-token charges. It now takes up to 400 questions per request (twice the previous limit), with US and EU regional processing. OpenAI says it is up to 10x faster than GPT-6 Luna through the Responses API. Why it matters: If you run classification, routing or risk-scoring steps on a general chat endpoint, this is worth benchmarking. The speed figure is OpenAI's own, so test it on your traffic. Source: openai.com/index/decisions-api 2. OpenAI models are coming to Google Cloud (Oct 21) What changed: OpenAI says its frontier models and Codex agentic workflows are coming to Google Cloud's Model Garden, sold through Google Cloud Marketplace, and companies with GCP spend commitments will be able to draw those down. GA is planned "in the coming weeks"; there's an access request form. OpenAI notes Microsoft remains its primary cloud partner. Why it matters: For GCP-first teams this could simplify procurement. No pricing or model list is published yet. Source: openai.com/index/openai-models-google-cloud 3. Google open-sources TorchTPU (Oct 20) What changed: TorchTPU, co-developed with Meta's PyTorch team, is on GitHub. Google says it runs existing PyTorch workloads on TPUs with minimal code changes, supports eager mode and torch.compile, and works with vLLM and SGLang. Why it matters: It lowers the cost of trying TPUs if GPU capacity is your bottleneck. Source: cloud.google.com/blog Would any of these change something in your stack? Reply with which one and why.
Sent using Claude