Meta Releases Muse Spark Open with 816B Parameters
Meta released Muse Spark Open, an 816B open MoE model with native multimodality and a 1M token context window built for agentic coding.
Nvidia GB300 NVL72 cuts AI inference costs by up to 7.2x
SemiAnalysis and vLLM log a 7.2x perf/TCO gain on Nvidia's GB300 NVL72 rack running MoE models, driven by copper interconnects and wide parallelism.
Perplexity Computer Adds Ultrafast Inference
Perplexity Computer adds ultrafast inference for GPT-6 Astra and SOL 6.1 on its Max tier, alongside a warning about rapid credit consumption.
Perplexity Computer Adds Live Polymarket Data
Perplexity Computer now pulls real-time Polymarket odds, volume, and liquidity across 250,000 prediction markets without executing bets.
Hermes Agent Connectors Add RBAC and Tool Guardrails
Nous Research released 50+ Hermes Agent connectors with RBAC controls, operational boundaries, and local execution that requires no account.
Waymo Opens Robotaxi Service Across Three US Cities
Waymo opened its robotaxi service to everyone in San Antonio, San Diego, and Tampa, ending waitlists and connecting San Antonio International Airport.
Fired OpenAI Safety Researchers Demand Inspectable AI
Three fired OpenAI safety researchers urge the industry to halt frontier models whose chain-of-thought reasoning cannot be inspected by human evaluators.
Inworld Inference API Pairs Fast Open Models with Voice
Inworld Inference launches with low latency for Gemma 4 and open models, bundling OpenAI-compatible completions with real-time text-to-speech.
OpenTelemetry in Claude: Anthropic Adds Native Traces
Anthropic rolled out native OpenTelemetry traces in Claude, exporting agent latency, tool calls, and token metrics directly to enterprise OTLP collectors.
Nvidia Rubin Delivers 4.2x Cost Efficiency in Inference
SemiAnalysis benchmarks show Nvidia Rubin running SGLang delivers 4.2x better cost efficiency on a 180B MoE model compared to Blackwell GB300 systems.