gatesfoundation
A workspace for the Gates Foundation DPI, FairFoward and Gooey teams focused on evals for low-resource languages plus the home of our Agriculture advisory work e.g. https://gooey.ai/ageval
2mo ago
Public
115 runs
113 runs
114 runs
Hausa audio benchmark run using Omnilingual ASR for transcription and Gemma 4 26B for the English response.
524 runs
Hausa audio benchmark run using Omnilingual ASR for transcription and Kimi K3 for the English response.
Hausa audio benchmark run using Intron ASR for transcription and Gemma 4 26B for the English response.
494 runs
Hausa audio benchmark run using Gemini 3.1 Pro Preview directly on audio input, with English-only responses.
462 runs
Hausa audio benchmark run using GPT-5.6 Terra directly on audio input, with English-only responses.
505 runs
(Updated Aug 5, 2026)This page presents a comparative demonstration of multiple LLM and speech-to-text systems combinations in Kannada → English translation pipelines. 
Use this workflow compare latency on Copilot Bulk Runs. The graph display the median score. Lower is better!
162 runs
81 runs