Latest AI Intelligence
Independent benchmarks, architectural audits, and empirical research.
-
Why Cheaper AI Models Cost More on 32% of Tasks as 89% of Enterprises Miss Spend Forecasts
Only 11% of 396 firms can forecast AI spend, while a 6,800-task Microsoft study proves cheaper models cost more on 32% of…
-
Red Hat Halves NVIDIA Nemotron 3.5 Lightning 30B Memory With FP8 for Single-GPU Agents
Red Hat AI released an FP8-quantized checkpoint of NVIDIA’s 30B Nemotron 3.5 Lightning Mamba-MoE model, cutting memory 50% to fit on a…
-
Anthropic Commits $100M to Claude Frontier Academy to Train 10,000 Deployed Engineers
Anthropic launched the $100M Claude Frontier Academy with Accenture and Bain to certify 10,000 Frontier Deployed Engineers on production AI agents.
-
Microsoft AI Releases MAI-Transcribe-2-Streaming, Topping 38 Speech Models With 2.5% Word Error Rate
Microsoft AI released MAI-Transcribe-2-Streaming, ranking #1 across 38 speech models on Artificial Analysis with 2.5% WER and 0.13s post-speech latency.
-
AWS Launches Strands Decider 2B: An Open-Weight Local Decision Model for AI Agents
AWS released Strands Decider 2B, an open-weight Qwen3.5-2B pointer model that validates AI agent tool calls in under 100ms on local GPUs…
-
Claude Opus 5.5 Intelligence, Performance and Price Analysis
Anthropic details Claude Opus 5.5 Intelligence, Performance. Comprehensive benchmarks, pricing tiers, and architecture audits.
-
Anthropic Offers Free Claude Cloud Credits of Up to $250
Anthropic grants $100 to $250 in promotional cloud credits to Claude Pro and Max subscribers to subsidize asynchronous Claude Code background execution.
-
SISTRIX Audit Exposes Stagnant Sources in AI Overviews
SISTRIX tracks 82,619 prompts over 17 weeks, finding 53% of Google AI Overviews reuse identical source domains while AI Mode rotates 56%…
-
SIGIR Benchmark Reveals Four Key Signals for AI Citations
A 252,000-trial SIGIR benchmark identifies topic fit, stated pricing, recency, and list order as primary drivers of AI citations, while formatting fails.
-
Randomized Study Measures 39.8% AI Overviews Click Loss
A randomized study finds Google AI Overviews cut outbound web clicks by 39.8%, proving generative summaries satisfy search intent without visiting sites.
-
Study Reveals How ChatGPT Retrieves and Caches Citations
A study of 1,200 ChatGPT answers reveals a 3-tier retrieval system where instant queries skip origin servers 93% of the time, relying…
-
OpenAI Releases GPT-6 Sol and Luna to Slash API Costs
OpenAI unveils GPT-6 Sol and Luna frontier models, cutting API inference costs by over 50% while advancing multi-modal reasoning and tool execution.