Research
Research and practical context
Follow the evidence behind models, infrastructure and the businesses developing AI.

Evaluations and recorded outputs
Record a model run locally and explore evaluation frameworks.
How long will a GPU keep doing useful work?
Evidence, launch dates and a buy-or-keep calculator.
What connects the GPUs?
A guide to interconnects, parallelism and cloud topology.
Practical guides
Hardware choices, memory requirements and deployment questions.
China Desk
Labs, model releases and the organisations behind them.
Emerging projects
New labs and architectures, with reported claims attributed.
How our comparisons work
Sources, memory estimates and API-price calculations.
What we are developing next
Our output recorder and evaluation-tool directory are available. OMHQ benchmark runs and documented deployment case studies remain in development. The current catalogue's memory figures are estimates, not speed or quality benchmarks.