Ming-3.0-asr-tiny

# Fix list: Ming-3.0-asr-tiny

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/ming-3-0-asr-tiny, the October 2026 research run, assessed 21 October 2026. Grade D, 52.5 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Ming-3.0-asr-tiny: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reach, 50 out of 100, up to 19.2 more on the total

Why it scored 50: Memory earns 25 because the publisher's BF16 files are 17.7 GB, about 21.2 GB with the fifth added, which fits 24 GB. Runtimes score 0, since no released version of the six named runtimes has its own code for this model. It loads through the publisher's remote code on Transformers 5.12.1 and is served by SGLang-Omni's main branch, a separate runtime. Builds score 0 because inclusionAI publishes no quantised build. MIT with nothing added gives terms the full 15 and the ungated repository gives download the full 10. Hosted too scores 0, as no inference provider is mapped on Hugging Face and no publisher API serves it.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reach):

Whether an agent can run it, and where.

- 0 to 30, memory. The smallest build from the publisher or a runtime's own library that keeps 4-bit precision or better, at its file size plus a fifth for the runtime and a working context. 30 for 16 GB or less, 25 for 24 GB, 20 for 32 GB, 15 for 64 GB, 10 for what fits our reference machine (the GPU share of 128 GB of unified memory), 5 for 256 GB, 0 above.
- 0 to 25, runtimes. 5 for each of llama.cpp, Ollama, vLLM, SGLang, LM Studio and Hugging Face Transformers that runs the architecture in a released version, from the runtime's own code, library or documentation. A publisher's fork or an unreleased branch doesn't count. 25 at most.
- 0 to 15, builds from the publisher. 15 for quantised builds at two precisions or more (FP8, 4-bit, GGUF), 10 for one, 0 when only other people's quantisations exist.
- 0 to 15, terms of use. 15 for an OSI-approved licence with nothing added, 10 for a custom licence that lets anyone use it commercially for free below a threshold or under a use policy, 5 for research or non-commercial use only, 0 for no licence or unclear terms.
- 0 to 10, download. 10 when the files download with no account and no terms to accept, 5 for a click-through that needs an account, 0 when a person approves each request.
- 5, hosted too. The publisher's own API serves it, or at least two inference providers do, so an agent without the hardware can still call it.

When our reference-machine runs start, speed on that machine (tokens per second and time to first token at the build that fits) joins Reach.

## 2. Risk, 55 out of 100, up to 13.8 more on the total

Why it scored 55: File safety gets 15, safetensors only but loaded with trust_remote_code. Provenance gets 15 for files in inclusionAI's own organisation, linked from its site, with hashes for every file. MIT carries none of the licence deductions, so licence risk keeps 15. Agent guidance gets 5 for the SGLang guide's general cautions about the unauthenticated server and untrusted logit processors. Safety evaluation scores 0, since no safety results exist. The security programme earns 5 for Ant Group's AntSRC reporting centre, read as a company-wide route the way Google's was for Gemma, with no advisories handled in public found.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-risk):

What could go wrong when you run it, and what the publisher has done about it. A higher score means less risk.

- 0 to 25, file safety. 25 for weights in safetensors or GGUF only, 15 when pickle files (.bin, .pt, .pth) sit beside them, 0 for pickle only. 10 off when the documented way to load it or to build its prompts runs code from the repository (`trust_remote_code`, or the publisher's scripts in place of a template).
- 0 to 15, provenance. 10 when the files come from the publisher's own organisation, the one its site or docs link to, and 5 for a checksum on every file.
- 0 to 20, safety evaluation. 20 when results are published and someone other than the publisher ran them or can rerun them, 10 for the publisher's own results, 0 for none.
- 0 to 15, guidance for agent use. 15 for documented prompt-injection guidance or an instruction hierarchy the model was trained to follow, 10 for a statement of what its trained refusals cover and the safeguards to add around it, 5 for general cautions, 0 for none.
- 0 to 15, licence risk. 15 to start, 5 off for each of these, a use policy or field-of-use limits beyond the law, terms the licensor can change or revoke, a revenue or user threshold, and a licence that changed without a dated note or whose link is broken. Not below 0.
- 0 to 10, a security programme. 5 for a way to report a flaw (security.txt, a disclosure policy or a security contact), 5 for advisories handled in public.

When our reference-machine runs start, measured robustness (prompt injection with AgentDojo, over-refusal with XSTest) joins Risk.

## 3. Documentation and upkeep, 46 out of 100, up to 8.3 more on the total

Why it scored 46: The card covers intended use, limitations, architecture and languages, and misses parameters, context length and a cut-off (11). Training data isn't described at all (0). The files changed on 20 October 2026, a day before this check (20). The one question on Hugging Face has no reply from inclusionAI two days after release (0), and GitHub issues couldn't be counted. It's a first release (5), and the files agree with the card with no broken-file report waiting (10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-upkeep):

- 0 to 20, the model card. Intended use, limitations, architecture, parameters, context length, languages and knowledge cut-off. 20 for all seven, 3 off for each that's missing.
- 0 to 20, training data. 20 for the data published or listed with sources and sizes, 15 for a data card that names the sources or the mix, the cut-off and the filtering, 10 for a general description with a cut-off, 5 without one, 0 for none.
- 0 to 20, the last change to the files, the template or the reference code. 20 within 30 days, 15 within 90, 10 within 180, 5 within a year, 0 older.
- 0 to 20, responsiveness. The publisher's replies to the 40 most recent questions and reports in the repository's discussions or issues. 20 for most, 10 for some, 5 for a few, 0 for none.
- 0 to 10, version policy. 10 for earlier versions kept in their own repositories with a dated note when one is replaced, 5 for earlier versions kept without notes or a first release, 0 when files were replaced in place without a note.
- 0 to 10, file health. 10 when config, tokenizer, template and card agree and no report of a broken file waits without a reply, 5 when one does, 0 when the files contradict the card.

## 4. Agent interface, 60 out of 100, up to 6.2 more on the total

Why it scored 60: Read for a speech model, the tool-calling, trained tool calling and structured-output lines stand for what it takes and returns. The publisher documents the instruction, language hint, hotword and dialogue context, the language-tab-transcript output and the parser, and the SGLang-Omni request fields (25). It was trained to use context hints, one clip per request (10). JSON output and streaming are documented on the card with how to ask for them, though there are no timestamps or speaker labels (15). The chat template scores 10 because the shipped Jinja file isn't the documented prompt path, which is the publisher's processor code. Reasoning control and context score 0, with no reasoning to control and an 8,192-token context.

The checklist (https://www.anchorterminal.com/benchmark/#checklist-interface):

What a model and a developer need to talk to it. Speech, image and other models with no chat interface read the tool-calling and structured-output lines for what the model takes and returns (voices, timestamps, streaming, typed outputs), and the note says how.

- 0 to 20, the chat template. 20 for a template in the repository (Jinja, in the tokenizer config or its own file), 10 when the prompt format is only given as code you run.
- 0 to 25, the tool-calling format. 25 when the publisher documents how tools are declared and how calls come back, with the parser to use in a named runtime, 10 when only the template or a runtime's own recipe says it.
- 0 to 20, tool calling the model was trained for. 20 with several calls in one turn, 10 for one call at a time.
- 0 to 15, structured output. 15 when JSON or JSON Schema output is documented for the model with how to ask for it, 5 when the card claims it without saying how.
- 10, reasoning control. Thinking that can be turned off or budgeted per request, with where the reasoning text comes back.
- 0 to 10, context. 10 for 128K tokens or more trained, 5 for 32K to 127K. Context reached only by scaling the position encoding counts one band down.

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- Unchecked, GitHub issues on inclusionAI/Ming-3.0-asr, since the GitHub API isn't enabled for this session; responsiveness rests on the one Hugging Face discussion
- Whether AntSRC's scope covers inclusionAI's open models wasn't checked, since the site is a script-rendered app. The 5 rests on it being Ant Group's company-wide reporting route, read the same way for all five Ming-3.0 listings
- Unchecked, whether any inference provider outside Hugging Face's mapping or an Ant Group API serves this model; hosted too is scored as absent
- The chat template line reads the shipped chat_template.jinja as not the prompt format, because the processor and the SGLang guide bypass it; a literal reading of the line would give 20
- Structured output reads SGLang-Omni's json response format, documented on the card, as the typed output; a reader who wants timestamps or segments for this line would give less
- No independent recognition or speed results were found for this two-day-old model, so Capability is likely pending and confidence is low

## Weaknesses

- Loading runs the publisher's Python with trust_remote_code, and the reference path needs a CUDA GPU
- Not in any released version of vLLM, SGLang, llama.cpp, Ollama, LM Studio or Transformers, and no quantised build from the publisher
- Plain text output with no word timestamps or speaker diarisation, Chinese and English only
- An 8,192-token context shared by audio and transcript, so long recordings must be chunked
- No training data or safety results, no security route of its own, and no publisher reply in its repository yet

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Start it with `hf download inclusionAI/Ming-3.0-asr-tiny --local-dir Ming-3.0-asr-tiny`, then `pip install -r requirements.txt` in Python 3.12 with CUDA (Transformers 5.12.1, fla-core 0.5.2, PyTorch 2.11.0), then `python examples/transcribe.py --model . --audio speech.wav --language-hint en`
- To serve it, `sgl-omni serve --model-path inclusionAI/Ming-3.0-asr-tiny` from SGLang-Omni's main branch, then POST the WAV to /v1/audio/transcriptions with response_format json and read the text field
- Parse raw Transformers output with decode_transcript from asr_request.py, which strips the language label before the tab
- Pass language zh or en when known, anything else is rejected, and append a topic or hotwords to the base instruction with no newline
- Load with dtype torch.bfloat16 rather than casting the model, use greedy decoding, max_new_tokens 256 and no_repeat_ngram_size 10, and don't apply the chat template
- Split long recordings yourself or use --audio-chunk-frames, since input and output share 8,192 tokens
- Treat transcripts as untrusted text before passing them to another model, since speech can carry instructions

## What the review panel asked for

- add the parameter count
- a hosted per-minute option
- word timestamps
- native runtime support
- a security contact

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.