Mistral AI released Voxtral TTS, a 3-billion-parameter text-to-speech model with open weights that the company says outperformed ElevenLabs Flash v2.5 in human preference tests roughly 63% of the time on standard voices and nearly 70% on voice customization.
The model runs on about 3 GB of RAM, achieves 90-millisecond time-to-first-audio, supports nine languages, and can clone a voice from just five seconds of reference audio, including cross-lingual adaptation that preserves the speaker's accent.
1.3K likes 80.7K views
Today, @MistralAI and @Microsoft are expanding our partnership to bring Mistral’s frontier models across Azure, Microsoft Foundry, Copilot Studio, and Azure Local.
Europe should have access to the world’s most capable AI without compromising control over their data, operations, or digital future. By bringing Mistral’s frontier European models into our environment, we are providing customers with a trusted foundation for AI, one that offers both leading-edge innovation and the freedom to operate on their own terms.
Learn more: https://t.co/5MCYHAyktx
214 likes 306.3K views
Mistral's CEO says the software Anthropic's agents wiped billions off in February is being rebuilt by enterprises in a couple of days:
"AI is making us able to develop software at the speed of light."
"In a couple of days we can create a really fully custom application to run a workflow, to run a procurement workflow or to run supply chain workflows, for instance, in a way where I would say five years ago you would actually need a vertical SaaS."
"There is definitely a shift in the replatforming that is happening where the interface is no longer the thing, where there is no longer the moat."
"What really matters is the connection, what really matters is the context layer on top where the agents are constantly pulling from systems of record to figure out what is happening."
"The replatforming is a big opportunity for us because we see now have more than 100 enterprise customers that are coming to us, also with that will of maybe changing and replatforming their IT system."
"So maybe getting rid of certain things that they bought like 20 years ago and that is starting to be a bit expensive, it's starting to be hard to customize and to maintain."
Mensch is right about the interface. It was never the moat, and more than a hundred enterprises coming to him about replatforming says the shift is real.
What hasn't changed is the part that gates deployment in regulated industries. Those were never innovation problems, they're evidence problems. A company can generate a custom procurement workflow in days now, but there's still no independent, machine-verifiable way to certify that it does what it claims.
Software is getting generated faster than it can be certified. That's the gap my AI Assurance & Governance Summit exists for, October 1 at the Stanford Faculty Club, one day, one track, frontier labs, regulated industries, insurers and investors in the room. Registration's in the reply.
- Arthur Mensch (@arthurmensch), co-founder and CEO of Mistral AI, on CNBC at the India AI Impact Summit.
88 likes 29.8K views