| --- |
| license: apache-2.0 |
| tags: |
| - cortiq |
| - cmf |
| --- |
| |
| # CMF — Cortiq Model Format |
|
|
| **One file. No Python, no torch, no CUDA install, no C++ toolchain.** |
|
|
| A `.cmf` carries the quantized weights, the tokenizer and the chat template |
| together, checks its own integrity, and memory-maps straight off disk. The |
| runtime is a small Rust core with no ML framework under it, running on CPU |
| everywhere and on GPU through wgpu — Vulkan, DX12, Metal — out of the box. |
|
|
| This repository is the format's front page. It holds the specification and |
| the index of published models; it carries no weights of its own. |
|
|
| - **Specification** — [`SPEC.md`](SPEC.md), the normative document |
| - **Source** — [github.com/infosave2007/cmf](https://github.com/infosave2007/cmf) (Apache-2.0) |
| - **Runtime** — [`cortiq-cli` on crates.io](https://crates.io/crates/cortiq-cli), and prebuilt binaries for six targets on the [releases page](https://github.com/infosave2007/cmf/releases/latest) |
|
|
| ## Run one |
|
|
| ```sh |
| cargo install cortiq-cli # or take a prebuilt binary |
| hf download infosave/Nanbeige4.2-3Bcmf nanbeige42-3b-q4t.cmf --local-dir . |
| cortiq verify nanbeige42-3b-q4t.cmf # per-tensor hashes |
| cortiq run nanbeige42-3b-q4t.cmf --prompt "Write a haiku about memory-mapped files." |
| ``` |
|
|
| Converting a checkpoint is one command and no Python: |
|
|
| ```sh |
| cortiq convert --model Qwen/Qwen3-0.6B --quant q8 --output qwen.cmf |
| ``` |
|
|
| ## What is actually in the file |
|
|
| A CMF file is an envelope, not a tarball: readers navigate through the header |
| and never guess at offsets. Unknown header fields are ignored, so the format |
| grows additively — a breaking change costs a feature bit or a version bump, |
| never a silent reinterpretation. A file written today stays readable, and |
| `cortiq verify` is the contract that says so. |
|
|
| Beside the weights it carries the things that are usually scattered across a |
| repository or lost entirely: the tokenizer and chat template, per-tensor |
| hashes, an optional Ed25519 signature, per-task **masks** that select an |
| active subset of the shared weights, and a **swarm of skills** sharing one |
| backbone — several specialists in one file, chosen at load time rather than |
| downloaded separately. |
|
|
| Two things it does that other single-file formats do not: |
|
|
| **Attention that stops growing with the context.** Converting with `--o1` |
| replaces a layer's softmax attention with a streaming operator holding a |
| fixed-size state — a few exact anchor keys, an exact recent window, and a |
| landmark sketch of everything older, under one shared denominator. The |
| weights are byte-identical; the flag only records a hint in the header. |
|
|
| **Generative models in the same container.** The format is not |
| LLM-only. `cortiq imagine` renders images and `cortiq animate` renders video |
| with synchronized stereo audio, from `.cmf` files packed the same way, on the |
| same runtime. |
|
|
| ## Published models |
|
|
| | model | what it does | size | |
| |---|---|---| |
| | [Nanbeige4.2-3B](https://huggingface.co/infosave/Nanbeige4.2-3Bcmf) | looped transformer, 22 layers run twice | 2.36 GB | |
| | [Bonsai-1.7B](https://huggingface.co/infosave/Bonsai-1.7Bcmf) | 1-bit BitNet, phone-sized | 0.33 GB | |
| | [Bonsai-8B](https://huggingface.co/infosave/Bonsai-8B_2bit_cmf) | 1-bit ternary at 2 bits stored | 2.32 GB | |
| | [Bonsai-27B](https://huggingface.co/infosave/Bonsai-27Bcmf) | 1-bit, 40 tok/s on an RTX 4090 | 5.10 GB | |
| | [KAT-Coder-V2.5](https://huggingface.co/infosave/KAT-Coder-V2.5-CMF) | 34.7B-A3B MoE for code | 12.65 GB | |
| | [Kimi-Linear-48B-A3B](https://huggingface.co/infosave/Kimi-Linear-48B-A3B-Code-CMF) | KDA linear attention, MoE | 17.75 GB | |
| | [Qwen3.6-27B](https://huggingface.co/infosave/Qwen3.6-27Bcmf) | dense, q4tp | 14.26 GB | |
| | [Qwen3.6-35B-A3B](https://huggingface.co/infosave/Qwen3.6-35B-A3Bcmf) | MoE, q4tp | 18.68 GB | |
| | [Qwen3.6-35B-A3B-Escha](https://huggingface.co/infosave/Qwen3.6-35B-A3B-Escha-W2cmf) | the same at two bits | 12.87 GB | |
| | [DeepSeek-V4-Flash](https://huggingface.co/infosave/DeepSeek-V4-Flash-0731-cmf) | 1T-A32B, split into parts | 455 GB | |
| | [Lumina-Image-2.0](https://huggingface.co/infosave/Lumina-Image-2.0cmf) | text → image, 19 GB tree in 3.2 GB | 6.72 GB | |
| | [MiniMax-H3 Turbo](https://huggingface.co/infosave/MiniMax-H3-Turbo-cmf) | text → video **with sound**, 4 steps | 71.05 GB | |
|
|
| Every one of them runs from the same binary: `cortiq run` for the text |
| models, `cortiq imagine` for Lumina, `cortiq animate` for MiniMax-H3. |
|
|
| ## Where it is honest |
|
|
| The **format** is settled at v2 and evolves additively. The **crate APIs** |
| may still move before 1.0. Quality claims in the model cards are measured |
| held-out numbers with the corpus named, not estimates — where a number was |
| not measured, the card says so rather than guessing, and the container |
| refuses to record a quality field without a measurement. |
|
|
