Model card: add Qwen3-ASR 0.6B (speech to text)

README.md

@@ -1,6 +1,8 @@
 ---
 license: apache-2.0
-base_model: Qwen/Qwen3.5-0.8B
+base_model:
+  - Qwen/Qwen3.5-0.8B
+  - Qwen/Qwen3-ASR-0.6B-hf
 tags:
   - gguf
   - llama.cpp
@@ -9,15 +11,19 @@ tags:
   - ios
   - macos
   - qwen3.5
+  - automatic-speech-recognition
+  - asr
+  - speech
+  - qwen3-asr
 pipeline_tag: text-generation
 language:
   - en
   - vi
 ---
 
-# DarkSurf models: small language models that run on phones
+# DarkSurf models: small language and speech models that run on phones
 
-Language models that have been measured and picked to **run on the device**, with no server: the apps in this
+Language and speech-to-text models that have been measured and picked to **run on the device**, with no server: the apps in this
 project (starting with DarkSurf, a privacy-first browser for macOS and iOS) download one of these files once and run it
 locally with [llama.cpp](https://github.com/ggml-org/llama.cpp) on Metal. Prompts and page text never leave the device.
 
@@ -29,6 +35,7 @@ out of this repo.
 | Catalog id | File | Size | Min. RAM | Use |
 |---|---|---:|---:|---|
 | `qwen3.5-0.8b-ud-q4kxl` | `Qwen3.5-0.8B-UD-Q4_K_XL.gguf` | 559 MB | 3 GB | Summarize, ask about a page, translate. Text only |
+| `qwen3-asr-0.6b-q8` (speech) | `Qwen3-ASR-0.6B-Q8_0.gguf` + `mmproj-Qwen3-ASR-0.6B-Q8_0.gguf` | 847 MB | 3 GB | Speech to text, 52 languages including Vietnamese |
 
 ### Qwen3.5-0.8B UD-Q4_K_XL
 
@@ -50,6 +57,28 @@ and at non-English text (Vietnamese translations read fluently but contain meani
 estimate, and speed and memory have **not yet been measured on an iPhone or iPad**. Perplexity was measured in English
 only.
 
+### Qwen3-ASR 0.6B Q8_0 (speech to text)
+
+[Qwen3-ASR-0.6B](https://huggingface.co/Qwen/Qwen3-ASR-0.6B-hf) (Apache-2.0), converted to GGUF here with llama.cpp
+`3c809f4` and quantized to Q8_0. **Two files, both required**: the language model (639 MB) and the audio encoder
+`mmproj-…` (208 MB); sizes and SHA-256 are in `catalog.json`. Converting the `-hf` checkpoint needs two fixes that
+llama.cpp's converter does not have yet (nested config and tensor names for the audio encoder, and the M-RoPE sections);
+the converted files match the official `ggml-org` GGUF tensor for tensor.
+
+Measured on an Apple M4 with llama.cpp `3c809f4` (Metal), on three short clips (English and Vietnamese synthetic speech,
+one LibriSpeech clip):
+
+| | Result |
+|---|---|
+| Word error rate | 1.8%, identical to the transformers fp32 reference (Vietnamese clip 0%) |
+| Speed | 40 s of speech in about 2 s after a 0.9 s load (about 20x faster than real time) |
+| Memory | 0.8 GiB peak footprint with a 4096-token context (1.4 GiB resident) |
+
+**Always set a small context (4096).** The default 65 536-token context needs 8 GiB for the KV cache alone.
+
+Limits: three clips are a smoke test, not an accuracy benchmark; noise, accents, long recordings and streaming are not
+measured, and **it has not been run on an iPhone or iPad yet**. The language label it prints is one per recording.
+
 ## Run it
 
 ```sh
@@ -61,12 +90,20 @@ llama-completion -m Qwen3.5-0.8B-UD-Q4_K_XL.gguf -c 8192 --top-k 20 --top-p 0.8
 The prompt is ChatML. The empty `<think></think>` block turns thinking off (faster, no scratchpad in the answer);
 without it Qwen3.5 starts with a long reasoning section. Tested with llama.cpp `3c809f4`.
 
+Speech to text (needs both files):
+
+```sh
+llama-mtmd-cli -m Qwen3-ASR-0.6B-Q8_0.gguf --mmproj mmproj-Qwen3-ASR-0.6B-Q8_0.gguf -c 4096 --temp 0 \
+  --audio speech.wav -p "Transcribe the audio."
+# -> language English<asr_text>Mr. Quilter is the apostle of the middle classes, ...
+```
+
 In an app, link llama.cpp as an xcframework (Metal) and load the file in-process with the same prompt and sampling
 settings. Keep the 8k context, and unload the model when idle.
 
 ## How the apps get models
 
-- `catalog.json` lists the models offered, each pinned to a commit of this repo (`/resolve/<commit>/…`) with its size
+- `catalog.json` lists the models offered (chat models in `models`, speech models with several files in `speech`), each pinned to a commit of this repo (`/resolve/<commit>/…`) with its size
   and SHA-256. `catalog.json.sig` is an Ed25519 signature; apps accept a downloaded catalog only if it verifies against
   the public key built into the app, and check every download against the catalog's size and SHA-256.
 - The file is downloaded when the user asks for it, and kept on the device until deleted.