LOFT · Hard retrieval; average context length up to 128K; 300 queries · Benchmark definition and scores
LOFT · Hard retrieval; average context length up to 128K; 300 queries has benchmark disclosures or scores across 1 model release.
Benchmark versions and metrics
Aliases LOFT (hard retrieval)
- Accuracy · % · Higher is better
Latest models by series · reported scores
The latest models in each named series have no scores to display in this scope. Recorded historical data appears below.
History · 1 release
- Gemini 2.5 Pro (GA)1 score record for this release, including 1 exact value (may include historical records)