MLC-SLM Challenge 2026 opens with $20,000 prize pool

By AI, Created 9:41 AM UTC, May 20, 2026, /AGP/ – Registration is open for the 2nd MLC-SLM Challenge 2026, a multilingual speech AI competition with a $20,000 prize pool and free access to a 2,100-hour dataset across 14 languages. The challenge is built to test systems for diarization, speech recognition and conversational understanding in real-world dialogue settings.

Why it matters: - The challenge gives researchers and companies a shared benchmark for multilingual speech AI. - The task set targets core capabilities for next-generation Speech LLMs: who is speaking, what was said, and what the conversation means. - Free access to a large dataset lowers the barrier to entry for academic teams, industry teams and individual researchers.

What happened: - The 2nd Multilingual Conversational Speech Language Models Challenge 2026, also known as the MLC-SLM Challenge 2026, opened for registration on April 30, 2026. - The competition offers a total prize pool of USD 20,000. - Registration is free. - The challenge invites academic teams, industry teams and individual researchers. - The organizing committee says the goal is to advance Speech Large Language Models for real-world multilingual conversational speech.

The details: - Registered participants get free access to a large-scale multilingual conversational speech dataset. - The dataset includes about 2,100 hours of two-speaker conversational speech across 14 languages. - The languages are English, French, German, Italian, Portuguese, Spanish, Japanese, Korean, Russian, Thai, Vietnamese, Tagalog, Urdu and Turkish. - The dataset also includes regional accents such as Canadian French, Mexican Spanish, Brazilian Portuguese, British English, American English, Australian English, Indian English and Philippine English. - Participants can enter two tasks. - Task 1 covers multilingual conversational speech diarization and recognition. - Task 1 requires systems to identify who is speaking when and transcribe multilingual conversational speech. - Task 1 evaluation will not provide oracle segmentation or speaker labels. - Task 2 covers multilingual conversational speech understanding. - Task 2 requires systems to use acoustic and semantic information to understand multilingual conversations. - Task 2 evaluation uses multiple-choice questions about the full conversation. - Task 2 tests meaning, context and speaker-level information. - Both pipeline-based systems and end-to-end Speech LLM systems are allowed. - External datasets and pretrained models are allowed if they are freely accessible and clearly reported. - The challenge is positioned as a benchmark for multilingual ASR, speaker diarization, Speech LLMs and spoken language understanding.

Between the lines: - The challenge builds on the first MLC-SLM Challenge, which drew 78 teams from 13 countries and regions. - The first edition recorded 489 valid leaderboard submissions and 14 technical reports. - The first challenge’s summary paper has been accepted by ICASSP 2026. - The expanded dataset suggests the organizers are trying to push evaluation closer to real-world multilingual speech conditions. - Allowing both pipeline and end-to-end approaches broadens participation and makes the competition relevant to multiple research strategies.

What’s next: - Registration remains open for participants. - The full competition details are available in the challenge announcement. - The organizers list mlc-slmw@nexdata.ai as the contact email for questions. - The registration form is available here. - The challenge team also directs participants to the competition page for more information.

The bottom line: - The MLC-SLM Challenge 2026 is a free, multilingual benchmark competition with cash prizes and a sizable speech dataset designed to test practical Speech LLM performance.