Google Cloud made daily token quotas generally available for BigQuery functions that invoke language models from SQL. AI.GENERATE, AI.GENERATE_TEXT, AI.CLASSIFY, AI.IF and related generation functions are covered. Administrators can set project and per-user ceilings to contain token use in large analytical queries.
An AWS guide dated October 12 explains serving a GGUF model through a SageMaker AI endpoint using the standard Deep Learning Container, returning OpenAI-compatible chat responses. It includes CPU and GPU deployment options.
AWS published a technical evaluation on October 9 of reinforcement-learning rollouts in Firecracker microVMs. It reported more than 100,000 samples across over 500 configurations using the labsweep measurement harness. In the tested workloads, tail latency rose sharply beyond one microVM per virtual CPU, while median latency changed much less.
Discovery AI converts plain-language questions into Sonar filters. AWS says the system cut query cost by 90%, input tokens by 93% and latency by 75% against its initial design.
FXC Intelligence announced an MCP server on October 15 connecting clients’ AI tools to cross-border payments benchmarks and market data. According to the company announcement published by AWS, it uses Amazon Bedrock and AgentCore, with dedicated agents for pricing, competitive analysis and go-to-market planning.
FXC Intelligence / AWS Press Center1 min read Data included
Microsoft described Cosmos DB vNext emulator updates for AI coding agents on October 15. Its report covers October releases’ binary-protocol support, stricter indexing-metadata validation and experimental Nacelle gateway, alongside tools for loading test data and collecting traces.
AWS detailed GPU kernel optimizations for mixture-of-experts models on SageMaker AI on October 13. The company described swapping matrix operands, routing FP8 operations to the Triton backend and removing unnecessary just-in-time recompilations.
AWS Artificial Intelligence Blog1 min read Data included
Microsoft AI said on October 15 that it completed pre-training MAI-Base-2, a mixture-of-experts model with 2.8 trillion total and 80 billion active parameters. The company says training used 16,000 GB200 GPUs over 38 days in August and September; the model is still undergoing post-training.
Global Emotions contains 525 hours of speech recorded by 10,420 speakers across 83 languages and 108 locales; the company made the dataset available for commercial and research use.