Phonon-2 is a 164 MB speech recognition model that matches the accuracy of a much larger teacher model while being 15 times smaller. It averages 5.21% word error across seven English test sets and runs extremely fast—transcribing an hour of audio in 20 seconds on a MacBook Air.

• Achieves 5.21% average word error rate across Open ASR Leaderboard benchmarks, beating larger models including Whisper large-v3-turbo (6.58%) and matching its 2.5 GB teacher model's accuracy despite being 1/15th the size
• Extreme speed: processes audio 174x faster than realtime on M5 MacBook Air, 143x on Zen 5 CPU, and 6,680x on H100 GPU in batches
• Quantized to 2.1 bits per weight using five learned levels, enabling the tiny download size while maintaining teacher-model performance

Microsoft released MAI-Transcribe-2, a speech recognition model priced at 10 cents per hour—down from 36 cents five months ago—that includes features like speaker identification and multi-language support that competitors charge extra for. The release signals Microsoft's broader strategy to build its own AI models and reduce dependence on OpenAI after restructuring their partnership.

• The pricing drops transcription costs low enough that it stops being a budget line item for enterprises; a bank processing 100,000 hours annually pays $10,000 instead of $36,000.
• Microsoft bundled premium features into the base product: speaker diarization, word-level timestamps, keyword biasing, code-switching between languages, and output modes for compliance versus readability—capabilities competitors sell separately.
• Three model releases in five months with expanding language coverage (25 to 60 languages) and improving rankings suggest Microsoft has stabilized the architecture and is now scaling aggressively to make transcription a commodity before competitors respond.
• The push toward in-house models reflects Microsoft's renegotiated OpenAI partnership, which ended exclusive access and revenue-sharing, freeing Microsoft to pursue its own frontier AI and cut costs across products like Teams, Word, and Excel.

Microsoft released MAI-Transcribe-2, a speech recognition model that claims to outperform competitors like OpenAI's Whisper and Google's Gemini on speed, accuracy, and cost. It handles 60 languages, includes speaker identification and word-level timestamps, and costs $0.10 per hour.

• 10x faster than GPT-Transcribe, 7x faster than Scribe v2, and 5x faster than Gemini 3.5 while maintaining higher accuracy across benchmarks
• Supports 60 languages with 5.2% average word-error rate on FLEURS benchmark, plus features like speaker diarization, keyword biasing, and code-switching for mixed-language conversations
• Priced at $0.10/hour as a limited-time launch offer, undercutting competitor pricing while maintaining competitive performance