Microsoft AI unveils seven models targeting lower-cost AI workloads

Microsoft AI says it is releasing seven new models designed to combine high capability with lower price-performance for developers and enterprises. The lineup targets coding, cybersecurity, image editing, real-time audio, transcription and decision-making workloads.
- Microsoft AI says it is releasing a family of seven new models.
- MAI Code 2 is claimed to reach 98% of GPT-6.1 Sol’s coding quality at about 40% of the price.
- MAI Realtime 1 starts at $0.01 per audio minute, according to Microsoft AI.
- Dragon Copilot Transcribe powers 10 million patient encounters a month, the company says.
- Maia 200 is reported to achieve 1.7x more tokens per watt.
Models target coding, cyber and multimodal tasks
Microsoft AI says MAI Cyber 2 Flash matches GPT-6 Astra’s quality at less than 2% of the cost per task on the CyberGymE2E benchmark. The company says MAI Code 2 reaches 98% of the coding quality of GPT-6.1 Sol at about 40% of the price, while reaching 96% performance compared with Opus 5.5 at 5% of the price.
MAI Code 2 Flash is described as matching the performance of GPT-6 Luna at 65% of the price and Haiku 5.5 at 20% of the price. Microsoft AI also says MAI Image 2.7 ranks first on Artificial Analysis image editing, outperforming GPT Image 2.6 Sunburst (max) at one fifth of the price.
Audio and decision models emphasize operating costs
MAI Realtime 1 is presented as outperforming GPT-Realtime 2.1 and starting at $0.01 per audio minute. Microsoft AI says Dragon Copilot Transcribe is built on MAI Transcribe 2, which it calls the world’s fastest and most accurate multilingual transcription model.
According to Microsoft AI, Dragon Copilot Transcribe already powers 10 million patient encounters a month, with multilingual error rates halved. Decision 1.1, the company’s latest decision model, is described as 50x faster and 79% cheaper than GPT-Sol 6.1.
Microsoft links model costs to its own AI hardware
The announcement also highlights Maia 200, Microsoft’s second-generation AI accelerator. Microsoft AI says the chip is now in production running its largest models and achieves 1.7x more tokens per watt.
The company frames the releases around rising token spending and the need for models that can run agents continuously on engineering work. For developers and enterprises, the stated price comparisons could make lower-cost options more relevant for sustained coding, audio and decision workloads. The performance figures are Microsoft AI’s claims and are tied to the benchmarks and model comparisons named in its announcement.
Microsoft AI: Seven new world class models at scalable price-performance | Microsoft AI


