Topic
Effective cost driven by caching, turn counts, retries and tool overhead rather than headline token prices.
The Cambridge company says routing agentic work across Nvidia, AMD and Google chips doubles accuracy at a quarter of the cost. No customer has confirmed those numbers.
Reality
Artificial Analysis puts OpenAI's new reasoning model at three times the index score of its price-tier peers and more than twice their token appetite. The output line dominates the invoice.
Reality
Arize and Fireworks ran ten models against 40 agent tasks and found the cheapest model per finished job also had the worst pass rate. Coverage, not price, is the binding constraint.
Reality
Cache-hit pricing went from 0.025 to 0.3 yuan, peak and off-peak rates arrived, and Moonshot, ByteDance and Alibaba are moving the same direction. Reprice now.
Reality
Stanford's Hazy Research group put a decomposed number on local inference efficiency. Hardware did more of the work than architecture, which changes who captures the gain.
Reality
Noreva expects US gas above $10 per million BTU at some hubs, against just under $3 at Henry Hub today. Fuel is roughly half the cost of power from a large plant.
Reality
Grok 4.6, Qwen3.8-Max and DeepSeek V4-Pro shipped inside about 24 hours, and two of the three came with downloadable weights. The benchmarks existed to justify a cheaper invoice.
Perspective Coverage
4 publishersReality