The Curious Engineer

All blogs / AI / ML

Mozilla.ai

blog.mozilla.ai

Open, transparent AI for real world impact. Built for developers, creators, and teams shaping what’s next.

AI / MLSoftware / engineering practiceDev tools / editors / IDEs

’24’25’26
Posts
93 from 23 Jan 2024 to 19 Oct 2026
Posting
weekly (56 in the last year)
Kind
company blog
Feed
https://blog.mozilla.ai/rss/

2026 48

  1. Who searches the web for your agent?
  2. Building the Missing Half
  3. Mistral Large 4 is live in Otari
  4. You Don’t Just Need Better AI. You Need Better Infrastructure.
  5. Benchmarking Local LLM Servers: llama.cpp, llamafile, LM Studio, and Ollama
  6. Introducing Agent Skills in Octonous
  7. Open Source Is Not a Virtue: It’s an Ownership Model
  8. Using Octonous as a Product Operations Manager
  9. llamafile v0.10.5
  10. The Future of AI is Personal — But Not The Way You Think
  11. How Frontier Labs Are Building Subtle Developer Lock-In
  12. Who Cares About LLM costs?
  13. Stop Chasing New Models. Build Once and Access Them All.
  14. From Evaluation to Guardrails: What We Brought to ACM FAccT 2026
  15. Open Models are ready for agents. Their APIs are not.
  16. Using Octonous as an AI Safety Engineer Discover how Octonous helps our team at Mozilla.ai cut through daily busywork, from automatically monitoring our libraries to staying on top of the latest developments in AI safety. By Daniel Nissani Jul 9, 2026
  17. The Control Layer: Why the Next Era of AI Is About Infrastructure, Not Just Models The Model’s the Easy Part - How to Get, and Keep, Value Here’s how I see the evolution of AI in enterprises over the last few years: * Autumn of 2022, the world thinks it’s going into a recession. IT budgets are froze
  18. Introducing Otari: The Open-Source LLM Control Plane If you are building LLM-powered applications today, you are probably managing multiple LLM providers, a pile of API keys, and your own logic for routing, budgets, and failovers. Accessing language models is no longer the challenge. Operating LLM i
  19. Announcing transcribe.cpp Meet transcribe.cpp, a new open-source C/C++ speech-to-text inference library with portable, GPU-accelerated support for multiple STT models. Developed through Mozilla.ai's Builders in Residence program, it makes adding fast, local transcription to applications easier than
  20. Using Octonous as a Product Manager A look at how we use Octonous inside mozilla.ai to reduce the everyday overhead of product work, from turning Slack feedback into GitHub issues to staying on top of product changes and finding context across the tools where work already happens. By Hasan Gözlügöl
  21. Image Classification Comes to encoderfile Encoderfile now handles images. Starting with image classification, you can run vision models as a single executable — no Python runtime, no serving infrastructure, just a file path in and a label out. By Javier Torres Ramón Jun 23, 2026
  22. What is an LLM control plane? Runaway agents? Provider outages? Discover why your AI stack needs an LLM control plane, not just a gateway, to handle production routing, budgets, and privacy. By Anushri Gupta Jun 12, 2026
  23. Technical Content Use the Otari Gateway with OpenCode AI coding sessions can feel like a black box. Route OpenCode through the Otari Gateway to track costs, token usage, and model activity in real time. Get budget controls and visibility across every session without changing a single line of applica
  24. Announcement Otari: Own Your AI Stack Meet Otari, an open-source LLM gateway powered by any-llm, and Otari.ai, the hosted platform built on the same foundation. Run frontier or open-weights models through one API with usage tracking, budget controls, routing policies, observability, and team managem
  25. Expert Opinion AI Got Expensive. Now What? Cloud AI pricing changed fast in 2026. This post looks at why more teams are moving back to local models, the tradeoffs behind tools like Ollama and LM Studio, and why portability and ownership are becoming bigger concerns for developers. By Anushri Gupta M
  26. Product Release cq exchange: Agents without Borders cq exchange gives agents a shared place to store and retrieve experience-driven knowledge through private namespaces and a public commons. By Anushri Gupta, Peter Wilson May 21, 2026
  27. Expert Opinion The Interface Is No Longer the Product The future of AI may not be agents using today’s apps. It may be apps rebuilt around structured representations agents can inspect, modify, and validate directly. The deck, doc, or dashboard becomes the output, not the source of truth. By Alejand
  28. Technical Content First Line of Defense for cq cq helps coding agents share resolution paths and learn from past failures. We partnered with Lauren Mushro to bring VIBE✓ into cq and help review knowledge units before they enter shared memory. By Lauren Mushro, Daniel Nissani May 12, 2026
  29. Octonous Open Beta: What We've Learned and Where We're Going The Octonous open beta is live. Learn what we discovered during closed beta, the workflow patterns users kept returning to, and the biggest improvements shipped since launch. By David de la Iglesia Castro, Khaled Osman, Hareesh Bahuleyan M
  30. Expert Opinion Sovereign AI: Control, Choice, and Why It Goes Beyond Geopolitics Sovereign AI shows up across nations, companies, communities, and individuals. This piece, based on a conversation with John Dickerson, CEO at Mozilla.ai, looks at control over AI systems, avoiding single points of fail
  31. Technical Content Encoderfile’s New Format: Why a “Dull” Design Wins Encoder models power most NLP in production, but deploying them still means dragging along Python runtimes and dependencies. Encoderfile introduces a single executable with an appended payload and a format that can be inspected and
  32. Announcement The Hardest Part of Running a Small Business in the Trades Running a small trade business includes a steady flow of admin work: quotes, scheduling, invoices, payments, and more. This post looks at how that workload builds up and introduces Clawbolt, a focused assistant designed around t
  33. Technical Content Hardening Your LLM Dependency Supply Chain When source code and distributed packages don’t match, risks increase. This breakdown of the LiteLLM incident shares what to watch for and how to reduce exposure. By Anushri Gupta Mar 25, 2026
  34. Announcement cq: Shared Agent Learning cq explores a shared commons where agents can query past learnings, contribute new knowledge, and avoid repeating the same mistakes in isolation. By Peter Wilson Mar 23, 2026
  35. Product Release llamafile Reloaded: What’s New in v0.10.0 llamafile 0.10.0 unifies portability and modern model features. Bundle weights, run multimodal models, and access tool calling and Anthropic Messages API support, all from a single executable. By Davide Eynard Mar 19, 2026
  36. When Shipping Becomes Too Easy AI is changing product development. When building becomes effortless, the real constraint is no longer code. It’s clarity, product judgment, and knowing when the right decision is not to ship yet. By Julie Belião Mar 17, 2026
  37. Federate Phishing Detection: Training a URL Classifier without Sharing Browsing Data Mozilla.ai joins Flower Hub as a launch partner with fed-phish-guard, a federated phishing detection project. The classifier trains across distributed clients and shares only model updates, allowing collaborative le
  38. Owning Code in the Age of AI AI lets engineers generate thousands of lines of code in minutes. But humans still reason about systems slowly. That gap forces a rethink of ownership, reliability, and where safety really lives in modern software systems. By Alejandro Gonzalez Mar 6, 2026
  39. Technical Content The Star Chamber: Multi-LLM Consensus for Code Quality The Star Chamber runs code reviews across multiple LLM providers and aggregates their feedback by consensus. Instead of relying on one model’s perspective, developers get a structured view of where models agree, disagree, and r
  40. Announcement any-llm in the Wild: Three Integrations as We Grow Our Ecosystem any-llm now integrates with JupyterLiteAI, LangChain, and Headroom. A single provider-agnostic layer powering notebooks, agents, and context optimization across OpenAI, Anthropic, Mistral, and local models. By Anushri Gupt
  41. Announcement Run OpenAI, Claude, Mistral, llamafile, and more from one interface, now in Go! Run any model, from any provider, like OpenAI, Claude, Mistral, or llamafile from one interface, now in Go. any-llm-go delivers type-safe provider abstraction, channel-based streaming, and normalized error h
  42. Technical Content Integrating Alinia into any-guardrail for Multilingual AI Security The newest integration with any-guardrail: Alinia AI, whose security models are specifically built to detect threats like prompt injection, data exfiltration, and policy violations by understanding the cultural and
  43. Technical Content Evaluating Multilingual, Context-Aware Guardrails: Evidence from a Humanitarian LLM Use Case A technical evaluation of multilingual, context-aware AI guardrails, analyzing how English and Farsi responses are scored under identical policies. The findings surface scoring gaps, reason
  44. Expert Opinion The Misunderstood Small Model Market: A Look Beyond OpenRouter's Data The State of AI report from OpenRouter and a16z offers valuable insight into API-based model usage. But many small models run locally on CPUs and consumer GPUs, outside managed services. This finding, however, warra
  45. Product Release Octonous: Making AI useful for everyday work Octonous helps people stay in flow while work runs across connected apps. The assistant takes real actions, reads messages, updates tools, and creates items while showing every step and asking for approval before anything changes. By Hasan
  46. Expert Opinion AI Generated Code Isn’t Cheating: OSS Needs to Talk About It Remember early 2025? "Vibe coding" was a meme and seemed mostly a tool for casual builders or those new to coding. It's now 2026, and we find ourselves living in a new reality. Industry leaders like DHH, Karpathy, and Lutke
  47. Product Release Secure Your Keys, Track Your Costs: any-llm Managed Platform Enters Open Beta Secure, encrypted LLM API key management across OpenAI, Anthropic, Google providers. Track costs, set budgets, avoid vendor lock-in. Free beta access now. By Alejandro Gonzalez, Dimitris Poulopoulos, Harees
  48. Product Release mcpd Plugins: Extend Your Agent Infrastructure Without Touching Your Code The new plugin system transforms mcpd from a tool-server manager into an extensible enforcement and transformation layer—where authentication, validation, rate limiting, and custom logic live in one governed pi

2025 32

  1. Announcement Building in the Open at Mozilla.ai: 2025 Year in Review The year 2025 has been a busy one at Mozilla.ai. From hosting live demos and speaking at conferences, to releasing our latest open-source tools, we have made a lot of progress and more exploration this year. By Kevin Tayong Dec 18,
  2. Technical Content Polyglot AI Agents: WebAssembly Meets the Java Virtual Machine (JVM) Leverage the JVM's polyglot capabilities to create a self-contained, enterprise-optimized server-side blueprint that combines the performance benefits of WebAssembly with the reliability and maturity of Java's eco
  3. Product Release Introducing any-llm managed platform: A secure cloud vault and usage-tracking service for all your LLM providers any-llm managed platform adds end-to-end encrypted API key storage and usage tracking to the any-llm ecosystem. Keys are encrypted client-side, never visible to us, while
  4. Product Release Encoderfile v0.1.0: Deploy Encoder Transformers as Single Binary Executables Encoderfile compiles encoders into single-binary executables with no runtime dependencies, giving teams deterministic, auditable, and lightweight deployments. Built on ONNX and Rust, Encoderfile is designed
  5. Product Release mcpd-proxy: Centralized Tool Access for AI agents in VS Code, Cursor, and Beyond With mcpd-proxy, teams no longer juggle multiple MCP configs. Run all servers behind one proxy and give every developer the same zero-config access inside their IDE. By Peter Wilson Nov 18, 2025
  6. Product Release Control LLM Spend and Access with any-llm-gateway Gain visibility and control over your LLM usage. any-llm-gateway adds budgeting, analytics, and access management to any-llm, giving teams reliable oversight for every provider. By Nathan Brake Nov 11, 2025
  7. Technical Content Can Open-Source Guardrails Really Protect AI Agents? AI Agents extend large language models beyond text generation. They can call functions, access internal and external resources, perform deterministic operations, and even communicate with other agents. Yet, most existing guardrai
  8. Product Release Run Any LLM with a Single API: Introducing any-llm v1.0 Run any model like OpenAI, Claude, Mistral, or llama.cpp from one interface. any-llm v1.0 delivers production-ready stability, standardized reasoning output, and auto provider detection for seamless use across cloud and local mo
  9. Product Release llamafile Returns Mozilla.ai is adopting llamafile to advance open, local, privacy-first AI—and we’re inviting the community to help shape its future. By Nathan Brake, Davide Eynard Oct 29, 2025
  10. Product Release Introducing any-guardrail: A common interface to test AI safety models Building AI agents is hard, not just due to LLMs, but also because of tool selection, orchestration frameworks, evaluation, safety, etc. At Mozilla.ai, we’re building tools to facilitate agent development, and we
  11. Technical Content 3W for In-Browser AI: WebLLM + WASM + WebWorkers What if AI agents could run entirely in your browser? Not just the UI part—the actual model inference, agent logic, and response generation, all happening locally without a single API call. By Baris Guler Aug 28, 2025
  12. Product Release Meet mcpd: requirements.txt for agentic systems mcpd is to agents what requirements.txt is to applications: a single config to declare, pin, and run the tools your agents need, consistently across local, CI, and production. By Peter Wilson, Alejandro Gonzalez, Hareesh Bahuleyan, Davi
  13. Expert Opinion Standardized Reasoning Content: A first look at using OpenAI’s gpt-oss on multiple providers using any-llm After OpenAI’s ChatGPT release, the default standard for communication was the OpenAI Completion API. However, with the new “reasoning models”, a critical piece of the output isn
  14. Product Release Introducing any-llm: A unified API to access any LLM provider When it comes to using LLMs, it’s not always a question of which model to use: it’s also a matter of choosing who provides the LLM and where it is deployed. Today, we announce the release of any-llm, a Python library that
  15. Technical Content Wasm-agents: AI agents running in your browser One of the main barriers to a wider adoption of open-source agents is the dependency on extra tools and frameworks that need to be installed before the agents can be run. In this post, we show how to write agents as HTML files, which c
  16. Technical Content Smarter Prompts for Better Responses: Exploring Prompt Optimization and Interpretability for LLMs Generative AI models are highly sensitive to input phrasing. Even small changes to a prompt or switching between models can lead to different results. Adding to the complexity, LLMs of
  17. Expert Opinion The Challenge of Choosing the Right LLM Imagine you could effortlessly navigate the universe of LLMs, always knowing which one is the perfect fit for your specific query. Today, this is a very difficult challenge. So, how do you efficiently manage and use LLMs for various tasks? This
  18. Technical Content What do you mean by AI testing...? We recently discussed the increasing need to test applications that make use of AI with tests that target problems specific to AI models. But an immediate follow-up question then arises: What specific problems? How is that testing different? By Ir
  19. Technical Content AIssert: Testing LLM Integrations Since LLMs exploded into public awareness, we have witnessed their integration into a vast array of applications. However, this also introduces new complexities, especially in testing. At Mozilla.ai, we did some research on the need to introduce fo
  20. Product Release Introducing Any-Agent: An abstraction layer between your code and the many agentic frameworks Since the launch of ChatGPT in 2022, generative AI and LLMs have rapidly entered everyday life. The viral adoption of these tools was unprecedented, and in some ways contentious. In order to
  21. Technical Content Evaluating Local LLMs on Translation Use Case with Lumigator LLMs have transformed a wide array of NLP applications, but concerns about privacy, data security, and control hinder LLM adoption for both individuals and enterprises. In this blog post, we focus on local LLMs, which off
  22. Product Release Open-source AI is hard. Blueprints can help! We’re excited to introduce Mozilla.ai Blueprints and the Blueprints Hub! Cut through the clutter of clunky tool integration, and see how you can speed up your development and spark your creativity. By Stefan French Mar 25, 2025
  23. Product Release 📘+🐊 Blueprint Releases & Lumigator's Latest Features Welcome to the first Mozilla.ai Update. We're thrilled to share the latest developments with you, including updates on our Blueprints and Lumigator toolkit. You'll also find community highlights and information about upcoming event
  24. Technical Content Build Your Own Timeline Algorithm: A Blueprint Timeline algorithms should be useful for people, not for companies. Their quality should not be evaluated in terms of how much time people spend on a platform, but rather in terms of how well they serve their users’ purposes. By Davide
  25. Technical Content Benchmarking DeepSeek R1 Models using Lumigator: A Practical Evaluation for Zero-Shot Clinical Summarization New state-of-the-art models emerge every few weeks, making it hard to keep up, especially when testing and integrating them. In reality, many available models may already me
  26. Technical Content Evaluating DeepSeek V3 with Lumigator A typical user may be building a summarization application for their domain and wondering: “Do I need to go for a model as big as DeepSeek, or can I get away with a smaller model?”. This takes us to the key elements: Metrics, Models, and Datase
  27. Technical Content Map Features in OpenStreetMap with Computer Vision Mozilla.ai developed and released the OpenStreetMap AI Helper Blueprint. If you love maps and are interested in training your own computer vision model, you’ll enjoy diving into this Blueprint. By David de la Iglesia Castro Mar 4,
  28. Technical Content Deploying DeepSeek V3 on Kubernetes Previously, we explored how LLMs like Meta’s Llama reshaped AI, offering transparency and control. We discussed open-weight models like DeepSeek and deployment options. Now, we show how to deploy DeepSeek V3, a powerful open-weight model, on a Ku
  29. Technical Content Running an open-source LLM in 2025 The landscape of LLMs has evolved dramatically since ChatGPT burst onto the scene in late 2022. At Mozilla.ai, we’re focused on improving trust in open-source AI by supporting their use in appropriate situations and their proper evaluation. By Nat
  30. Technical Content Structured Question Answering When deciding on a new Blueprint, we focus on selecting an end application that is both practical and impactful, along with the best techniques to implement it. With endless possible applications of LMs today, selecting one that is actually useful can
  31. Product Release Lumigator is here! Lumigator is a developer-first tool designed and built by the community to help engineers evaluate and compare AI models with ease. Lumigator empowers developers to make data-driven choices when integrating AI models into their applications. By Juliana Rodrigues Ar
  32. Technical Content Blueprint Deep Dive: Turn Documents into Podcasts Locally with Open-Source AI Blueprints are customizable workflows that help developers build AI applications using open-source tools and models. In this blog, we’ll dive into our first Blueprint: document-to-podcast. We’ll explain h

2024 13

  1. Product Release Introducing Blueprints: Customizable AI workflows for Developers Developers today face many challenges when trying to integrate AI into their apps or building an “AI solution” from scratch. At Mozilla.ai, we’re committed to breaking down these barriers with Blueprints – our initiativ
  2. Technical Content Let’s build an app for evaluating LLMs Lumigator 🐊 is a self-hosted, open-source Python application for evaluating large language models using offline metrics. It targets common machine learning use-cases, starting with summarization, and is extensible at the task and job level. By
  3. Technical Content Taming randomness in ML models with hypothesis testing and marimo The behavior of ML models is often affected by randomness at different levels, from the initialization of model parameters to the dataset split into training and evaluation. Thus, predictions made by a model are pote
  4. Product Release Introducing Lumigator 🐊 With hundreds of models, each offering different capabilities, the process can be overwhelming. That’s why Mozilla.ai is developing Lumigator, a product designed to help developers confidently select the best LLM for their specific project. By Juliana Rodrigue
  5. Expert Opinion The Importance of Ground Truth Data in AI Applications: An Overview We are witnessing explosive growth in the development of artificial intelligence (AI) applications across various industries, from virtual agents to healthcare diagnostics and autonomous driving. By Sandra Antunes Sep
  6. Expert Opinion Towards Truly Multilingual AI: Breaking English Dominance At the TAUS Massively Multilingual AI conference in Rome, I had the honor of moderating a panel on Multilingual AI provocatively titled “No English Please - How to move towards truly multilingual AI”. In this panel, we explored
  7. Technical Content Uncovering GenAI Trends: Using Local Language Models to Explore 35 Organizations Over the past few months at Mozilla.ai, we’ve engaged with a number of organizations to learn how they are using language models in practice. By Stefan French Jun 3, 2024
  8. Technical Content On model selection for text summarization We want to help users by providing an easy way for them to a) evaluate whatever model they choose to use, b) get understandable and transparent results, and c) trust the tech stack they want to build. In short, we’re offering accessibility,
  9. Technical Content Local LLM-as-judge evaluation with lm-buddy, Prometheus and llamafile In the bustling AI news cycle, where new models are unveiled at every turn, cost and evaluation don’t come up as frequently but are crucial to both developers and businesses in their use of AI systems. It is well
  10. Expert Opinion Open Source in the Age of LLMs Since our start last year, our key focus has been exploring state-of-the-art methods for evaluating and fine-tuning large-language models (LLMs), including continued open-source LLM evaluation experiments and establishing our GPU cluster’s infrastructure
  11. Technical Content The cost of cutting-edge: Scaling compute and limiting access In a year marked by extraordinary advancements in artificial intelligence, largely driven by the evolution of large language models (LLMs), one factor stands out as a universal accelerator: the exponential growth in comp
  12. Technical Content LLM evaluation at scale with the NeurIPS Large Language Model Efficiency Challenge ‍After a year of breakneck innovation and hype in the AI space, we have now moved sufficiently beyond the peak of the hype cycle to start asking a critical question: are LLMs good enough yet to solve
  13. Announcement Introducing Mozilla.ai: Investing in trustworthy AI Announcing Mozilla.ai, a startup building a trustworthy and open-source AI ecosystem with agency, accountability, and openness at its core. Mozilla will make an initial $30M investment in the company to fund the vision for making it ea