Mozilla

Post-quantum computing isn’t in the future. It’s our reality now. Its power is definitely exciting. Unless you’re talking data encryption. Then it’s a bit terrifying, because the encryption protecting all that sensitive information… …isn’t yet ready for all that quantum power. That power gap matters. The technology that stores and safeguards our sensitive data today runs a real risk of becoming overwhelmed and outsmarted by quantum computing tomorrow. This puts government communications, financial information, health records, and all sort of data we expect to remain private at risk. And no company or country can close that gap alone. It’s a topic that Mozilla will tackle this Global Encryption Day. On October 21, Mozilla’s Neha Kochar and Svea Windwehr join the European Commission and ENISA for Getting Ready for Q-Day: Post-Quantum Encryption Leaves the Lab — looking at what Firefox and others are already doing, where readiness is falling behind, and what needs to happen before the threat catches up.

One size does not fit all. Not in t-shirts, and certainly not in AI. Right now, a handful of huge AI systems are being built for scale. Bigger + faster = best. But Mozilla Foundation is betting big on small: Human-Scale AI, built for the people and problems it's meant to serve. Applications for our new Incubator cohort are open now, with up to $500,000 available per project over three years. Are you or someone you know imagining and building people-first tech alternatives? Apply today💡 link in comments

Can't make it to MozFest this year? You can still participate! Join MozFest's two-week online hackathon where builders can work on their side-projects with Mozilla.ai's Otari, an open source LLM router! Alejandro González Pérez and Marcus Rein will share details of how you can participate in the livestream. Come with questions!

Everyone’s building bigger. Mozilla Foundation is backing the builders doing something different. Keep your eyes peeled for something small arriving on October 21st 👀 🚀

Open source isn’t just about AI. It's about ownership.  Picture a future dominated by accessible, local AI running on personal devices, where foundational contributions and open collaboration are at the forefront. Getting there responsibly takes three things: full control of your models, transparency into how they work and the freedom to pick the model that works best for you. If you missed our live, the full video will be on Mozilla’s Youtube channel. Stay tuned for more information about the next Owners Not Renters show! ft: Marcus Rein, Alejandro González Pérez, Thomas Bille and Brian Mbuba

What’s the most vital asset for an AI company today? While funding and revenue matter, trust is another critical asset. Mozilla is supporting the launch of the Trustworthy AI Map, a market map of 79 VC-backed companies across 15 countries. It covers seven layers of the AI stack, including identity, data control, interoperability and evaluation. For this to work, trust has to be baked into the product or the business model through human agency and aligned business models. The map connect products with the companies and the capital behind them that allows them to explore commercial opportunities. Over the past year, we’ve invested through Mozilla Ventures, launched Otari, published The State of Open Source AI report, and partnered with Mila - Quebec Artificial Intelligence Institute to help organizations own and run their AI locally. The Trustworthy AO Map is another example of how Mozilla builds technology that people and businesses can trust. https://lnkd.in/g4aTaMDq

Decisions about today’s tech are made without us. A small group of companies control the money, the infrastructure, and the distribution. And when one future looks inevitable…well, we stop believing alternatives are possible. But the future isn’t decided yet.  And Mozilla Foundation is ready to bet that bigger isn’t always better. More to come, October 21⚡

the $15 model that bills like $31. Kimi K3 lists at $15 per million output tokens. on the artificial analysis intelligence index it spends 130M output tokens against a 63M median. 67M above the median. the effective rate is about $31. same suite. same list rate. $1,950 in output instead of $945. per-token price is the wrong unit for agentic. nearly every published comparison still uses this metric, yet it hides a 2x spread in what a task actually costs.  verbosity is trainable. in one post-training pass, DeepSeek’s V4-Flash-0731 cut output tokens 12% (234M → 206M) and gained 10 index points. the 28M tokens it dropped were never buying the score. it’s still 3.3x the median, it just moved down a tier. so does verbosity pay? on a median-verbosity workload at list rates: Fable 5 $3,150, K3 $1,950, Opus 5 $1,575. K3 beats one and loses to one. on measured cost per index task: K3 $0.86, Opus 5 $2.03, Fable 5 $3.15. it beats both. Opus doesn’t run at median verbosity either. before you commit to a model, compare what each task costs you.  full data in V1.1 of State of Open Source AI. link in the comments.

Today we’re announcing something new: Mozilla is moving into the enterprise browser market with a new product and business, Firefox Enterprise. Aimed initially at organizations in Europe that want more control over their technology infrastructure and data, Firefox Enterprise combines our browser with a management console that customers operate on their own infrastructure. IT teams can manage browser policies, access and data flows themselves. Our bet? The Growing international concern over dependence on a handful of powerful technology providers will create demand for an alternative, one built on Gecko, Mozilla’s independently developed browser engine. Institutions are questioning who controls the technology they rely on. Firefox Enterprise is our answer to that challenge, grounded in an idea we’ve held for 25 years: alternatives matter. Learn more about Firefox Enterprise in the first comment below. It’s a significant expansion for us, turning technology we’ve spent decades building for the consumer web into the foundation of a new enterprise business.

We backed Nous Research early. Now we’re doubling down. Congrats to the Nous team on their $90M raise to bring Hermes, their open-source AI agent, to businesses. More than 22 million downloads since February show that people want AI they can actually own, inspect and build on. Open source isn’t a side project for us. It’s how we think the internet (and AI) should work: transparent, accountable and in the hands of the people who use it. Proud to keep building that future alongside Nous.