Model card: all four models, per-folder licences (Apache 2.0 and LFM 1.0), sources, tags and the TestFlight link

README.md

@@ -1,51 +1,83 @@
 ---
-license: apache-2.0
-base_model: openbmb/MiniCPM5-1B
+license: other
+license_name: apache-2.0-and-lfm1.0
+license_link: https://huggingface.co/publicarray/fieldnote-models#licences
+base_model:
+- openbmb/MiniCPM5-1B
+- openbmb/MiniCPM5-2B
+- Qwen/Qwen3.5-2B
+- LiquidAI/LFM2.5-1.2B-Instruct
+base_model_relation: quantized
 pipeline_tag: text-generation
 library_name: coreai
 tags:
-  - core-ai
-  - ios
-  - on-device
-  - minicpm5
+- core-ai
+- coreai
+- apple
+- ios
+- iphone
+- on-device
+- neural-engine
+- minicpm5
+- qwen3.5
+- lfm2.5
 ---
 
 # Fieldnote models
 
-MiniCPM5 1B compiled ahead of time for Apple's Core AI, one ready-to-load bundle per
-iPhone chip family. Used by [Fieldnote](https://github.com/array-ai/fieldnotes), an
-on-device meeting recorder, to write meeting notes without compiling the model on
-the phone (the on-phone compile of a larger model ran an iPhone 16 out of memory).
+Small language models compiled ahead of time for Apple's Core AI, one ready-to-load
+bundle per iPhone chip family. Used by [Fieldnote](https://github.com/array-ai/fieldnotes),
+an on-device meeting recorder, to write meeting notes on the phone. A phone with a
+matching chip downloads a compiled bundle instead of compiling the model itself,
+which ran an iPhone 16 out of memory for the larger models.
+
+**Try Fieldnote:** [join the TestFlight beta](https://testflight.apple.com/join/c2M82CMX)
+(iPhone 15 Pro or later, iOS 27).
 
 ## Bundles
 
-| Folder | Core AI architecture | Phones |
-|---|---|---|
-| `ios-h17g` | h17g | iPhone 16 family |
-| `ios-h17p` | h17p | iPhone 16 family |
-| `ios-h18p` | h18p | iPhone 17 Pro |
+| Folder | Model | Chip | Phones | Licence |
+|---|---|---|---|---|
+| `ios-h17g` | MiniCPM5 1B | h17g | iPhone 16 family | Apache 2.0 |
+| `ios-h17p` | MiniCPM5 1B | h17p | iPhone 16 family | Apache 2.0 |
+| `ios-h18p` | MiniCPM5 1B | h18p | iPhone 17 Pro | Apache 2.0 |
+| `minicpm5-2b/ios-h17p` | MiniCPM5 2B | h17p | iPhone 16 family | Apache 2.0 |
+| `minicpm5-2b/ios-h18p` | MiniCPM5 2B | h18p | iPhone 17 Pro | Apache 2.0 |
+| `qwen3.5-2b/ios-h17p` | Qwen3.5 2B | h17p | iPhone 16 family | Apache 2.0 |
+| `qwen3.5-2b/ios-h18p` | Qwen3.5 2B | h18p | iPhone 17 Pro | Apache 2.0 |
+| `lfm2.5-1.2b/ios-h17p` | LFM2.5 1.2B Instruct | h17p | iPhone 16 family | LFM Open License 1.0 |
+| `lfm2.5-1.2b/ios-h18p` | LFM2.5 1.2B Instruct | h18p | iPhone 17 Pro | LFM Open License 1.0 |
 
 Each folder is a complete bundle: `metadata.json` (pointing at the compiled model),
-`tokenizer/`, and `minicpm5_1b_minicpm5_pal8_g32_static.<arch>.aimodelc`. A phone
-needs only the folder whose architecture matches `AIModel.deviceArchitectureName`.
-Other devices should use the portable model below and let Core AI compile it.
+`tokenizer/`, and the compiled `.aimodelc`. A phone needs only the folder whose chip
+matches `AIModel.deviceArchitectureName`. Other devices should use the portable
+export (below) and let Core AI compile it.
+
+## Sources
+
+Every bundle is a portable Core AI export compiled for one chip. No weights were
+changed.
 
-## Source and build
+| Model | Original | Portable Core AI export, pinned revision |
+|---|---|---|
+| MiniCPM5 1B | [openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B) | [mlboydaisuke/MiniCPM5-1B-CoreAI](https://huggingface.co/mlboydaisuke/MiniCPM5-1B-CoreAI) `ios-static/` @ `f38d143ec50cb2e54d8388479b919e3b9283943b` |
+| MiniCPM5 2B | [openbmb/MiniCPM5-2B](https://huggingface.co/openbmb/MiniCPM5-2B) | [mlboydaisuke/MiniCPM5-2B-CoreAI](https://huggingface.co/mlboydaisuke/MiniCPM5-2B-CoreAI) `ios-static/` @ `39db5ff9480e5d80e423889e3bacb7f9e9b9e40f` |
+| Qwen3.5 2B | [Qwen/Qwen3.5-2B](https://huggingface.co/Qwen/Qwen3.5-2B) | [mlboydaisuke/qwen3.5-2B-CoreAI](https://huggingface.co/mlboydaisuke/qwen3.5-2B-CoreAI) `gpu-pipelined-b2/qwen3_5_2b_decode_int8hu_block32_sym` @ `14e77014db5ae2f3de94825165f5ae58bededd7d` |
+| LFM2.5 1.2B | [LiquidAI/LFM2.5-1.2B-Instruct](https://huggingface.co/LiquidAI/LFM2.5-1.2B-Instruct) | [mlboydaisuke/LFM2.5-1.2B-CoreAI](https://huggingface.co/mlboydaisuke/LFM2.5-1.2B-CoreAI) `gpu-pipelined-b2/lfm2_5_1_2b_instruct_decode_int8hu_block32_sym` @ `3e3241cbcf368a663768dcc4e5785c4f0a02bae4` |
+
+## Build
 
-- Model: [openbmb/MiniCPM5-1B](https://huggingface.co/openbmb/MiniCPM5-1B), 8-bit
-  palettized, 4,096-token context.
-- Portable Core AI export: [mlboydaisuke/MiniCPM5-1B-CoreAI](https://huggingface.co/mlboydaisuke/MiniCPM5-1B-CoreAI)
-  `ios-static/` at revision `f38d143ec50cb2e54d8388479b919e3b9283943b`.
-- Compiled on GitHub Actions (Xcode 27, Metal Toolchain) by
-  [`compile-models.yml`](https://github.com/array-ai/fieldnotes/blob/main/.github/workflows/compile-models.yml):
+Compiled on GitHub Actions (Xcode 27, Metal Toolchain) by
+[`compile-models.yml`](https://github.com/array-ai/fieldnotes/blob/main/.github/workflows/compile-models.yml):
 
 ```
-xcrun coreai-build compile minicpm5_1b_minicpm5_pal8_g32_static.aimodel \
-  --platform iOS --min-deployment-version 27.0 --architecture <arch> --output <dir>
+xcrun coreai-build compile <model>.aimodel \
+  --platform iOS --min-deployment-version 27.0 --architecture <chip> --output <dir>
 ```
 
-Compute units were left to Core AI's default choice. Some on-device specialization
-still happens at first load, but not the full compile.
+The MiniCPM5 1B bundles leave the compute unit to Core AI. The Qwen3.5, LFM2.5 and
+MiniCPM5 2B h18p bundles are compiled for the Neural Engine. Some on-device
+specialisation still happens at first load, but not the full compile.
 
 ## Loading
 
@@ -53,16 +85,25 @@ still happens at first load, but not the full compile.
 import CoreAI
 import CoreAILanguageModels
 
-let arch = AIModel.deviceArchitectureName           // e.g. "h17g"
-let bundle = modelsDirectory.appending(path: "ios-\(arch)")
+let chip = AIModel.deviceArchitectureName           // e.g. "h17p"
+let bundle = modelsDirectory.appending(path: "qwen3.5-2b/ios-\(chip)")
 let model = try await CoreAILanguageModel(resourcesAt: bundle)
 ```
 
 Fieldnote downloads only the matching folder, from a pinned revision of this repo,
 and checks every file against its SHA-256 before use.
 
-## Licence
+## Licences
+
+This repo holds models under two licences. Each model's folder carries its own
+`LICENSE`:
+
+- **MiniCPM5 1B and 2B** (OpenBMB) and **Qwen3.5 2B** (Qwen team, Alibaba Cloud):
+  [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). The 1B's licence
+  is the `LICENSE` at the repo root; the others are in `minicpm5-2b/` and `qwen3.5-2b/`.
+- **LFM2.5 1.2B Instruct** (Liquid AI): LFM Open License v1.0, in `lfm2.5-1.2b/LICENSE`.
+  It allows commercial use only by organisations with under US$10 million in annual
+  revenue; read it before using these bundles.
 
-MiniCPM5 is released by OpenBMB under the Apache License 2.0, and so are these
-compiled bundles (see `LICENSE`). No weights were changed; the files here are the
-portable export compiled for specific chips.
+Core AI exports by mlboydaisuke; ahead-of-time compiled builds by publicarray. The
+files here are modified from the originals only by being compiled for a specific chip.