Urdu Evaluation of Audio AI Models Urdu Evaluation of Audio AI Models By Gooey.AI, CLEAR Global and the Gates Foundation • 28 Apr 2026 • Gooey.AI/language 0 5 10 15 20 25 30 35 0.4 0.5 0.6 0.7 0.8 Accuracy Latency Lower is better More accurate GPT 5.5 — Accuracy 0.63, Latency 24.27GPT 5.5 GPT 5.5 + Omni — Accuracy 0.72, Latency 25.3GPT 5.5 + OmniMost Accurate Gemini 3.1 Pro — Accuracy 0.71, Latency 22.27Gemini 3.1 Pro Gemini 3.1 Pro + Omni — Accuracy 0.55, Latency 32.38Gemini 3.1 Pro + Omni Gemini 3.1 Flash-Lite — Accuracy 0.43, Latency 6.19Gemini 3.1 Flash-LiteFastest Gemini 3.1 Flash-Lite + Omni — Accuracy 0.72, Latency 9.55Gemini 3.1 Flash-Lite + OmniMost Accurate Kimi K2.6 + Omni — Accuracy 0.47, Latency 22.2Kimi K2.6 + Omni GPT-OSS 120B + Omni — Accuracy 0.58, Latency 11.88GPT-OSS 120B + Omni GPT-Realtime 1.5 — Accuracy 0.45, Latency 11.21GPT-Realtime 1.5 GPT-Realtime 1.5 + Omni — Accuracy 0.69, Latency 15.92GPT-Realtime 1.5 + Omni Gemma 4 26B + Omni — Accuracy 0.61, Latency 10.91Gemma 4 26B + Omni GPT 5.5 — Accuracy 0.63, Latency 24.270 GPT 5.5 + Omni — Accuracy 0.72, Latency 25.31 Gemini 3.1 Pro — Accuracy 0.71, Latency 22.272 Gemini 3.1 Pro + Omni — Accuracy 0.55, Latency 32.383 Gemini 3.1 Flash-Lite — Accuracy 0.43, Latency 6.194 Gemini 3.1 Flash-Lite + Omni — Accuracy 0.72, Latency 9.555 Kimi K2.6 + Omni — Accuracy 0.47, Latency 22.26 GPT-OSS 120B + Omni — Accuracy 0.58, Latency 11.887 GPT-Realtime 1.5 — Accuracy 0.45, Latency 11.218 GPT-Realtime 1.5 + Omni — Accuracy 0.69, Latency 15.929 Gemma 4 26B + Omni — Accuracy 0.61, Latency 10.9110 © 2026 Gooey.AI