A curated set of news blending advanced AI, cost optimisation and infrastructure updates, highlighting practical opportunities for software-driven businesses.
GPT‑6.1 Terra: Faster, visual and computer‑using AI
OpenAI introduced GPT‑6.1 Terra, a model that blends detailed visual reasoning and computer‑use capabilities with more natural writing. Its standard API price is 50 % lower than GPT‑5.6 Terra, making it practical for high‑volume automation such as invoice reading and data entry.
Companies can lower automation costs for document‑heavy workflows by switching to Terra in existing AI pipelines.
How Dexpert helps: Artificial IntelligenceSources
- OpenAI: Introducing GPT‑6.1 Terra
LLM price drops of up to 80 % from OpenAI and Anthropic
OpenAI introduced the Luna ($1 / $6 / $0.10) and Terra models, while Anthropic released Claude Sonnet 5 and Claude Opus 5 with pricing between $2 / $10 and $5 / $25 per million tokens, amounting to up to 80 % reductions versus prior releases.
Enterprises can reconsider scaling AI workloads, opting for the cheaper models for high‑frequency tasks such as chatbots or ticket classification.
How Dexpert helps: Artificial Intelligence
Open‑weight AI models now cheap and competitive on‑prem
IBM's CEO stated that open‑weight models deliver 90 % of the performance of proprietary models at just 10 % of the cost, running on two or three server racks.
Businesses requiring on‑prem AI for latency or compliance can shift to open models, cutting infrastructure and licensing expenses.
How Dexpert helps: Artificial IntelligenceSources
Generative Polly and TorchNeuron for Trainium
AWS introduced a generative engine for Amazon Polly, enabling creative voice synthesis, and launched TorchNeuron, a library that optimizes PyTorch model training on Trainium chips.
Voice‑app developers can deliver richer experiences with Polly, and AI teams can lower training costs by leveraging TorchNeuron on AWS infrastructure.
How Dexpert helps: Cloud & Serverless
pg_ivm 1.17 adds materialized view maintenance enhancements
Version 1.17 of pg_ivm introduces a replace option for restore_immv(), pg_upgrade support, fixes an incorrect populated state, and improves unique‑key derivation for automatic index creation.
Teams relying on materialized views can streamline upgrades and improve index consistency, lowering operational overhead and risk in PostgreSQL deployments.
How Dexpert helps: Backend & APIsSources
- PostgreSQL Global Development Group: pg_ivm 1.17 released
Want to put one of these to work in your company?
Dexpert builds custom systems, apps, integrations and AI solutions. Tell us what you need and get an estimate, or take the free digital maturity assessment.