The Qeravio AI Change Index
A rolling view of verified AI change, built from public evidence rather than search volume, social attention or vendor claims.
Measured August 27, 2026, UTC. Rolling 30-day window.Where change is concentrated
infrastructure16
agents-software10
models-research1
Regional distribution
global26
europe1
Event types
releases27
Evidence trail
Latest events included in the index
The 4.54.0 version of the software makes function call output call IDs optional.
Importance 95, confidence 100.
Sorbet structured-output models added in v0.82.0
Importance 95, confidence 100.
The 2026-08-27 release of v3.54.0 makes function call output call IDs optional.
Importance 95, confidence 100.
Llama adds TENSOR_READ_LAZY to model loader
Importance 95, confidence 100.
Hugging Face Xet downloads now reuse tokens, reducing API calls per file.
Importance 95, confidence 100.
Llama-server and llama-cli gain benchmark-only synthetic speculative acceptance.
Importance 95, confidence 100.
Apple patches Metal memory leaks in macOS and iOS.
Importance 95, confidence 100.
Llama adds CPU-offload option for dense FFN weights
Importance 95, confidence 100.
What enters the index
- The event must be public, current and backed by a verified editorial candidate.
- The source must be active, rights-approved and visible in the evidence chain.
- The verification confidence must be at least 90 and the event must pass Qeravio's indexability gate.
- Counts are descriptive. They do not measure market size, investment value or product quality.
The page closes itself to indexing when the minimum evidence threshold is not met. Historical values are not reconstructed or guessed.