[ { "url": "https://api.github.com/repos/modular/modular/releases/390871966", "assets_url": "https://api.github.com/repos/modular/modular/releases/390871966/assets", "upload_url": "https://uploads.github.com/repos/modular/modular/releases/390871966/assets{?name,label}", "html_url": "https://github.com/modular/modular/releases/tag/max/v26.6.0", "id": 390871966, "author": { "login": "Ahajha", "id": 44127594, "node_id": "MDQ6VXNlcjQ0MTI3NTk0", "avatar_url": "https://avatars.githubusercontent.com/u/44127594?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Ahajha", "html_url": "https://github.com/Ahajha", "followers_url": "https://api.github.com/users/Ahajha/followers", "following_url": "https://api.github.com/users/Ahajha/following{/other_user}", "gists_url": "https://api.github.com/users/Ahajha/gists{/gist_id}", "starred_url": "https://api.github.com/users/Ahajha/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Ahajha/subscriptions", "organizations_url": "https://api.github.com/users/Ahajha/orgs", "repos_url": "https://api.github.com/users/Ahajha/repos", "events_url": "https://api.github.com/users/Ahajha/events{/privacy}", "received_events_url": "https://api.github.com/users/Ahajha/received_events", "type": "User", "user_view_type": "public", "site_admin": false }, "node_id": "RE_kwDOJcq7QM4XTDue", "tag_name": "max/v26.6.0", "target_commitish": "max/v26.6", "name": "MAX 26.6 / Mojo 1.1.0", "draft": false, "immutable": false, "prerelease": false, "created_at": "2026-09-17T16:06:35Z", "updated_at": "2026-09-17T16:14:51Z", "published_at": "2026-09-17T16:14:51Z", "assets": [ ], "tarball_url": "https://api.github.com/repos/modular/modular/tarball/max/v26.6.0", "zipball_url": "https://api.github.com/repos/modular/modular/zipball/max/v26.6.0", "body": "In MAX v26.6, we've added support for [audio generation](https://max.modular.com/serve/audio-generation/) with MiniMax-Music3, which can generate 44.1 kHz stereo music from a text prompt. We've also added support for Inkling from Thinking Machines, a multimodal model that accepts text and image inputs with NVFP4 mixture-of-experts, and GLM-5.3, plus expanded coverage across the Kimi, Qwen, and Gemma model families. Speculative decoding now supports DFlash and DSpark draft models on Gemma 4 31B, which speeds up token generation without changing the model's output. We've also sped up kernels across NVIDIA and AMD GPUs, including up to 4.8× faster Gemma 4 decode attention, 7.9× faster MoE routing, 16× faster low-batch NVFP4 quantization on B200, and up to 6.6× faster decode attention projections on MI355. Plus, the [`max.gpu`](https://max.modular.com/api/mojo/max/gpu/) package now includes everything previously provided in Mojo's `std.gpu` package, making it a complete entry point for accelerator programming.\r\n\r\nWith Mojo 1.1, we've added support for inferred member references—a leading-dot form now resolves against the expected type (for example, `SIMD[.float64, 4]` instead of `SIMD[DType.float64, 4]`). We've improved compile time and generated code performance, and the language server now suggests fixes for common off-by-one typos. The move from legacy closures to unified closures continues—more APIs take closures as runtime arguments. We've marked more APIs in the standard library as stable, such as `String`, `SIMD`, and `List`. Some deprecated items are now removed, such as `fn`, `alias`, and `__comptime_assert` keywords and the `@parameter if` and `@parameter for` syntax. And much more.\r\n\r\nPlus, we've recently opened more of the Mojo development process: we're now accepting contributions to the Mojo compiler!\r\n\r\nFor many more details about each package update, check out the full [MAX release notes](https://max.modular.com/releases/v26.6/) and [Mojo release notes](https://mojolang.org/releases/v1.1.0/).", "reactions": { "url": "https://api.github.com/repos/modular/modular/releases/390871966/reactions", "total_count": 28, "+1": 8, "-1": 0, "laugh": 0, "hooray": 8, "confused": 0, "heart": 7, "rocket": 3, "eyes": 2 } }, { "url": "https://api.github.com/repos/modular/modular/releases/368691898", "assets_url": "https://api.github.com/repos/modular/modular/releases/368691898/assets", "upload_url": "https://uploads.github.com/repos/modular/modular/releases/368691898/assets{?name,label}", "html_url": "https://github.com/modular/modular/releases/tag/max/v26.5.0", "id": 368691898, "author": { "login": "Ahajha", "id": 44127594, "node_id": "MDQ6VXNlcjQ0MTI3NTk0", "avatar_url": "https://avatars.githubusercontent.com/u/44127594?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Ahajha", "html_url": "https://github.com/Ahajha", "followers_url": "https://api.github.com/users/Ahajha/followers", "following_url": "https://api.github.com/users/Ahajha/following{/other_user}", "gists_url": "https://api.github.com/users/Ahajha/gists{/gist_id}", "starred_url": "https://api.github.com/users/Ahajha/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Ahajha/subscriptions", "organizations_url": "https://api.github.com/users/Ahajha/orgs", "repos_url": "https://api.github.com/users/Ahajha/repos", "events_url": "https://api.github.com/users/Ahajha/events{/privacy}", "received_events_url": "https://api.github.com/users/Ahajha/received_events", "type": "User", "user_view_type": "public", "site_admin": false }, "node_id": "RE_kwDOJcq7QM4V-cq6", "tag_name": "max/v26.5.0", "target_commitish": "max/v26.5", "name": "MAX 26.5 / Mojo 1.0.0", "draft": false, "immutable": false, "prerelease": false, "created_at": "2026-08-11T15:13:55Z", "updated_at": "2026-08-11T16:38:07Z", "published_at": "2026-08-11T16:05:56Z", "assets": [ ], "tarball_url": "https://api.github.com/repos/modular/modular/tarball/max/v26.5.0", "zipball_url": "https://api.github.com/repos/modular/modular/zipball/max/v26.5.0", "body": "In this release, we've moved most of the GPU programming APIs out of the Mojo standard library and into a new top-level `max` package—`std.algorithm` is now `max.algorithm`, `std.benchmark` is `max.benchmark`, and `std.gpu.{compute,host,memory,sync}` is `max.gpu.*`, along with the Mojo `layout` package. All of which is now part of the new [MAX accelerator library](https://docs.modular.com/api/mojo/), which also includes the existing AI kernels. We've also extended Apple silicon GPU support back to M1 devices, and added hardware-MMA flash-attention prefill for M5 to speed up TTFT. The MAX model bring-up workflow also gains new agent skills in [our skills repo](https://github.com/modular/skills), including new `serve-model`, `benchmark-model`, `eval-model`, and `profile-model` skills. We've also expanded model coverage and capabilities, made significant strides speeding up the Python API's experimental eager interpreter, sped up various kernels across NVIDIA, AMD, and Apple GPUs, and much more to accelerate models served with MAX.\r\n\r\nOf course, this release also brings Mojo 1.0! 🔥\r\n\r\nWith Mojo 1.0, we've started marking standard library APIs as stable, which means they won't be removed or changed in ways that break compatibility—we started with a small set that we'll grow in subsequent 1.x releases. There are so many changes in this release, it's difficult to summarize. It includes lambdas, pointer unification, interior origins, required `var` declarations, constraint improvements, faster Python interop, and so much more. We worked hard on getting names, defaults, and safety boundaries right for 1.0. That does mean you'll find more breaking changes than usual in this release. But nearly every breaking change ships with a deprecated alias and a compiler fix-it, so migration is mechanical.\r\n\r\nCheck out all the updates with the full [MAX release notes](https://docs.modular.com/releases/v26.5/) and [Mojo release notes](https://mojolang.org/releases/v1.0.0/).", "reactions": { "url": "https://api.github.com/repos/modular/modular/releases/368691898/reactions", "total_count": 67, "+1": 23, "-1": 0, "laugh": 0, "hooray": 28, "confused": 0, "heart": 5, "rocket": 11, "eyes": 0 } }, { "url": "https://api.github.com/repos/modular/modular/releases/341516225", "assets_url": "https://api.github.com/repos/modular/modular/releases/341516225/assets", "upload_url": "https://uploads.github.com/repos/modular/modular/releases/341516225/assets{?name,label}", "html_url": "https://github.com/modular/modular/releases/tag/max/v26.4.0", "id": 341516225, "author": { "login": "Ahajha", "id": 44127594, "node_id": "MDQ6VXNlcjQ0MTI3NTk0", "avatar_url": "https://avatars.githubusercontent.com/u/44127594?v=4", "gravatar_id": "", "url": "https://api.github.com/users/Ahajha", "html_url": "https://github.com/Ahajha", "followers_url": "https://api.github.com/users/Ahajha/followers", "following_url": "https://api.github.com/users/Ahajha/following{/other_user}", "gists_url": "https://api.github.com/users/Ahajha/gists{/gist_id}", "starred_url": "https://api.github.com/users/Ahajha/starred{/owner}{/repo}", "subscriptions_url": "https://api.github.com/users/Ahajha/subscriptions", "organizations_url": "https://api.github.com/users/Ahajha/orgs", "repos_url": "https://api.github.com/users/Ahajha/repos", "events_url": "https://api.github.com/users/Ahajha/events{/privacy}", "received_events_url": "https://api.github.com/users/Ahajha/received_events", "type": "User", "user_view_type": "public", "site_admin": false }, "node_id": "RE_kwDOJcq7QM4UWx_B", "tag_name": "max/v26.4.0", "target_commitish": "max/v26.4", "name": "MAX 26.4 / Mojo 1.0.0b2", "draft": false, "immutable": false, "prerelease": false, "created_at": "2026-06-18T15:56:24Z", "updated_at": "2026-06-18T17:05:06Z", "published_at": "2026-06-18T16:15:54Z", "assets": [ ], "tarball_url": "https://api.github.com/repos/modular/modular/tarball/max/v26.4.0", "zipball_url": "https://api.github.com/repos/modular/modular/zipball/max/v26.4.0", "body": "The MAX 26.4 release brings Apple silicon GPU serving to almost all compatible MAX models, with M3 and newer devices now running Llama and Qwen model families (and others that fit in system memory). We've also added support for new models, including Tencent Hunyuan Hy3-preview, LiquidAI LFM2, and GLM-5/5.1 with FP8/NVFP4 weights. DeepSeek V2 and V3 can now run on a single GPU (with block-scaled FP8 quantization for V3) and long-context sparse MLA for DeepSeek-V3.2. Gemma 4 gains native FP8 attention with an FP8 KV cache on B200 (roughly matching bf16 accuracy while doubling KV cache capacity) and NVFP4 quantization, structured output with grammar enforcement, MTP speculative decoding, and multi-GPU support for the 31B variant. MiniMax-M2 picks up MXFP4 quantization and tensor-parallel attention combined with expert-parallel MoE (TP+EP). Plus much, much more, including enhanced Prometheus metrics, improved kernel performance, new benchmark datasets, and lots of Python API upgrades.\r\n\r\nThe Mojo 1.0 beta 2 (v1.0.0b2) release makes steady progress towards a stable 1.0 release later this year. It removes the requirement that collection elements be `Copyable`: `List`, `Deque`, `LinkedList`, `InlineArray`, `Dict`, and `Set` now accept move-only elements. Trailing `where` clauses are now supported on struct declarations and `comptime` alias declarations. `String` and `StringSlice` gain Unicode-aware subscripts. The `fn` function declarations are now a hard compilation error (completing the `def`/`fn` unification begun in beta 1). Implicit `std` imports are also an error, making all standard-library dependencies explicit and freeing common names like `algorithm` and `memory` for user modules. On the CLI side, `mojo package` renamed to `mojo precompile` and the `.mojopkg` extension renamed to `.mojoc` (which is also a smaller, faster format). And that's just the start of it.\r\n\r\nFor much more detail about the MAX and Mojo changes, see the [MAX changelog](https://docs.modular.com/max/changelog/#v264-2026-06-18) and [Mojo changelog](https://mojolang.org/releases/v1.0.0b2/).", "reactions": { "url": "https://api.github.com/repos/modular/modular/releases/341516225/reactions", "total_count": 6, "+1": 2, "-1": 0, "laugh": 0, "hooray": 1, "confused": 0, "heart": 0, "rocket": 3, "eyes": 0 } } ]