benchmarks.site

Data

CSV data

One file per result table, exactly as shown on the site.

ASR: Polish speech recognition

Diarization: who is speaking

Omni models: video and audio input

OCR: handwritten notes, boards and textbook pages

OCR: exam-sheet photos with handwritten answers

Language models: classification, summaries, chat with tools

Language models: short answers, grading text, time and cost

Reasoning level (reasoning_effort): what models actually do

Embeddings: retrieval accuracy in Polish

Platforms: availability, latency and limits

Platforms: errors, reasoning billing and prices

Malware detection: AMTSO, EICAR containers, Objective-See hashes, Mac malware families

Malware detection: test samples vs held-out clean files

Malware scanning: static-heuristic false alarms before and after fixes

ML classifiers on TLSH fingerprints of Mach-O files

Scanner performance: database loading, re-scans, TLSH in SQL, Aho-Corasick