MLC-SLM Challenge 2026 opens, $20K pool

This is more valuable as a benchmark drop than a flashy launch: the real draw is the dataset and evaluation design, not the prize money. For speech LLM teams, it targets the hard parts current systems still struggle with, especially speaker changes and conversation-level understanding.

• –Free entry and free data lower the barrier for both academic labs and industry teams.
• –The 14-language mix plus regional accents makes this a real multilingual stress test, not an English-first ASR benchmark.
• –The multiple-choice understanding track is a useful check on semantic comprehension, not just transcription quality.
• –Diarization plus recognition is where a lot of speech systems still break, so strong results here should be meaningful.
• –Nexdata is positioning the challenge as both a research benchmark and an ecosystem builder around speech data.