Modular ships Mojo 1.0 as a stable systems language

Modular announced that Mojo, the language it has been developing since 2023, has reached version 1.0 as part of the company's 26.5 release. Modular frames 1.0 as the point where Mojo stops being a fast-moving internal project and becomes a stable foundation developers can build long-term projects on. The company says it now relies on Mojo every day in production as the foundation of its own commercial infrastructure, MAX and Modular Cloud.

Much of the 26.5 release finishes a round of language cleanup aimed at removing duplicate ways of expressing the same idea. Variables are now consistently declared with var, closures have been unified, and there is now a single Pointer type, alongside a number of renamings meant to make Mojo's vocabulary more precise. On top of that cleanup, 26.5 adds new capabilities: Python-style lambda syntax for inline closures, a more stable and reliable Mojo LSP server for editors such as VS Code, memory-safety diagnostics that catch reference invalidation (for example, flagging when List.append invalidates an existing reference into the list), and more consistent use of where clauses across the standard library, now able to carry a descriptive message when a constraint fails. Modular also says its Mojo AI Skills, covering tasks like new project creation, GPU programming, and porting code from other languages, are now considered "1.0 ready."

Modular is explicit that 1.0 is not the end of Mojo's evolution. During the 1.x line, the company says changes should primarily be additive, and while breaking changes may still happen, they will be managed the way mature languages such as C++ handle change over time. Planned future work includes a robust asynchronous programming model and pattern matching with unions, among other capabilities listed on Mojo's public roadmap. Modular also reaffirmed a commitment to progressively open-source more of the language and of the MAX components built with it, with a specific promise to open-source the Mojo compiler and toolchain in 2026, though no month was given.

Modular credits its community for the milestone alongside its own team. Since the standard library was open-sourced, nearly 200 contributors have landed more than 1,100 pull requests, changing over 200,000 lines of code, while more than a thousand other people have filed issues that helped shape the language.

The same release brings changes to MAX, Modular's model-serving platform. Installation is being split into extras so users install only what they need: max["serve"] and max["benchmark"] (or max-serve and max-benchmark with conda) pull in narrower dependency sets, while max["all"] installs everything; the older, all-in-one modular package will be retired in the 26.6 release. MAX 26.5 also adds support for two new hybrid Mamba-2 model families, GLM-5.2 and Nemotron-H, and Kimi 2.5 now works with Module V3, Modular's streamlined model-authoring path. Modular's open source collection of agent skills, used internally to speed up model lifecycle work, has picked up more than 7,200 downloads through skills.sh.

Modular is pointing developers to install or upgrade with uv pip install --upgrade mojo and uv pip install max[all], and says it will share more about its plans for Mojo, MAX, and further open-sourcing at ModCon on August 18 in San Francisco, available both via livestream and an in-person waitlist.

Key facts

Mojo reaches 1.0 as part of Modular's 26.5 release, its first stable, production-ready milestone since launching in 2023.
Since the standard library was open-sourced, nearly 200 contributors have landed more than 1,100 pull requests, changing over 200,000 lines of code, and more than a thousand others have filed issues.
26.5 unifies syntax (var-only variable declarations, a single Pointer type, unified closures) and adds features including Python-style lambda syntax and diagnostics for reference-invalidation memory-safety bugs.
MAX 26.5 adds support for the GLM-5.2 and Nemotron-H model families, splits installation into max["serve"]/max["benchmark"]/max["all"] extras, and retires the older modular package in 26.6.
Modular commits to open-sourcing the Mojo compiler and toolchain in 2026 and says it will detail further plans at ModCon on August 18 in San Francisco.

Why it matters

Mojo 1.0 is Modular's declaration that a language it has iterated on rapidly since 2023 is now safe to build long-term projects on, not just a moving target. The company backs that claim with its own usage: it says Mojo already underpins its commercial infrastructure, MAX and Modular Cloud, in production. For a language pitched at high-performance code across CPUs, GPUs, and accelerators, a stability guarantee matters more than another feature release, since it is what lets library authors and companies commit to Mojo without expecting their code to break on the next update.

Who it affects

Developers and companies building on Mojo or on Modular's MAX platform, including anyone maintaining Mojo libraries or tools who previously had to track frequent language changes. It also affects users of MAX's model support directly, since 26.5 adds the GLM-5.2 and Nemotron-H model families and brings Kimi 2.5 onto MAX's Module V3 authoring path. The roughly 200 contributors and 1,000-plus issue filers who have engaged with Mojo's open source standard library are the community Modular credits for reaching this point.

How to use it

Install or upgrade with uv pip install --upgrade mojo and uv pip install max[all]. MAX installation is now split into extras: max["serve"] and max["benchmark"] (max-serve and max-benchmark under conda) pull in only the dependencies needed for those tasks, while max["all"] installs everything; the previous all-in-one modular package is being retired in the 26.6 release. Modular's Mojo AI Skills, covering new project creation, GPU programming, and porting from other languages, are now marked 1.0 ready, and its open source collection of agent skills is available through skills.sh.

How solid is it

The claims come from Modular's own blog post announcing the release, so the production-readiness framing and community statistics (nearly 200 contributors, 1,100-plus pull requests, over 200,000 lines changed, more than a thousand issue filers) are self-reported rather than independently verified. The company does point to concrete, checkable facts to back the milestone: its own reliance on Mojo for MAX and Modular Cloud in production, and specific numbers around standard library contributions. No third-party benchmarks or independent reviews are included in the announcement itself.

Risks and caveats

Modular's stability promise is qualified, not absolute: changes during the 1.x line "should primarily be additive," but the company explicitly says breaking changes may still happen, managed the way mature languages like C++ handle change over time. The Mojo compiler and toolchain are not yet open source; Modular commits only to doing so "in 2026," without naming a month. As with any vendor announcement, the framing of Mojo as production-ready and the community metrics cited are Modular's own characterization of its project.

“During the 1.x timeframe, changes should primarily be additive, giving developers confidence that the language will not continually shift beneath them.”

— Modular, Mojo 1.0 announcement