MLX-VLM is a Python package for running inference and fine-tuning vision-language models, which process images and text together, on Mac computers using MLX. It is used by developers working with multimodal models and includes command-line, web, server, scripting, and training workflows.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add Blaizzy/mlx-vlm --skill hf-cache-modelsgit clone --depth 1 https://github.com/Blaizzy/mlx-vlmWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/blaizzy/mlx-vlm/hf-cache-models)<a href="https://agentmods.dev/skills/blaizzy/mlx-vlm/hf-cache-models"><img src="https://agentmods.dev/badge/skills/blaizzy/mlx-vlm/hf-cache-models/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/blaizzy/mlx-vlm/hf-cache-models"><img src="https://agentmods.dev/badge/skills/blaizzy/mlx-vlm/hf-cache-models.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00038 | $0.00525 |
| Opus 5.5 | $0.00015 | $0.00210 |
| Sonnet 5.5 | $0.00008 | $0.00105 |
| Haiku 4.5 | $0.00004 | $0.00052 |
Grade A, and why
hf-cache-models scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
curl http://127.0.0.1:8080/v1/models How it starts
The opening of the file, as written. The whole thing — 61 lines — stays where its author put it; the contents beside it link to each section on GitHub.
HF Cache Models
Query a running MLX-VLM server to list model candidates and their loaded status.
Use its configured URL and authentication; the examples assume the default local
address. See Skill("mlx-vlm-skills:server-inference") for server setup.
List Models
curl http://127.0.0.1:8080/v1/models
The JSON response's data array contains model IDs and a loaded boolean.
Both /models and /v1/models include loaded models, the shared Hugging Face
cache, and configured custom directories by default.
Supported Model Rule
The server checks discovered candidates with mlx_vlm/server/model_discovery.py:
- repo type is
model - readable model metadata exists in
config.jsonor a pipelinemodel_index.json - nonempty safetensors weights exist, including every shard referenced by any weight index
- all cached revisions are checked, preferring
main; another revision is returned using its absolute snapshot path - tokenizer metadata is not required
Discovery does not load weights, execute checkpoint Python, or prove generation works. Architecture compatibility is checked when loading a model.
Custom Model Directories
Each path can be a model folder or a parent containing model folders as immediate
children. Configure the server with repeated --model-dir options, or
MLX_VLM_MODEL_PATHS with paths separated by os.pathsep (: on macOS/Linux).
For directories scoped to one API request, use repeated model_dir query
parameters on /models or /v1/models. Paths are on the server's filesystem and
are added to its configured directories without persisting them:
curl --get http://127.0.0.1:8080/v1/models \
--data-urlencode 'model_dir=/Volumes/Models' \
--data-urlencode 'model_dir=/Users/me/my custom model'
Reporting
When reporting the result, include:
- server URL and any custom directories requested
- number of model candidates
- exact model IDs
- which models have
loaded: true
If this becomes part of a bug report, switch to Skill("mlx-vlm-skills:reproducible-github-issues").
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago Changed · -5 lines · -22 tokens per session 7004900611b3
- 1mo ago First seen · 66 lines · 60 tokens per session scan A 12c61061e272
hf-cache-models is a skill published in the GitHub repository Blaizzy/mlx-vlm (5,648 stars, last pushed today), licensed MIT. It adds 38 tokens to every session and 525 once invoked, about $0.0002 per session on Opus 5.5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
mlx-model-porting
Guides and validates architecture-aware ports of PyTorch/Hugging Face models to Apple MLX, inspects existing local MLX projects, and plans evidence-gated optimizations for Apple Silicon. Use when the user asks to run, port, convert, inspect, quantize, benchmark, or fix a model (LLM, VLM, audio/TTS/ASR, diffusion, SSM…
exporting-to-fhir
Convert OpenMed NER output (entities from openmed.analyzetext) into FHIR R4 resources — Condition, MedicationStatement, Observation — using OpenMed's built-in FHIR R4 export helpers in openmed.clinical.exporters. Covers the verified CodeableConcept builder (coding, codeableconcept, systemuri), deterministic fullUrl…
batch-processing-clinical-text
Run large-scale batch NER, PII extraction, or de-identification over many clinical notes on-device with OpenMed, with sharding, checkpointing, resumability, and append-only JSONL output. Use when the user needs to process a corpus or folder of notes, de-identify a dataset, run NER over thousands of documents, build a…
bridging-presidio-and-spacy
Combine OpenMed clinical NLP with Microsoft Presidio, spaCy, or LangChain through OpenMed's built-in interop adapter registry (openmed.interop). Covers the lazy adapter registry (availableadapters, getadapter, adapterspec), the presidio/spacy/langchain pip extras, and the verified callables — Presidio…
choosing-openmed-models
Discover and pick the right OpenMed model for a clinical or biomedical task, domain, or language. Use when the user asks which OpenMed model to use, wants to list model categories, find a Disease vs Oncology vs Privacy/PII model, get a PII model for a specific language, search models by size or task, or inspect a…
extracting-clinical-entities
Run clinical and biomedical named-entity recognition on medical text with OpenMed's analyzetext. Use when the user wants to extract diseases, drugs, anatomy, genes, or other biomedical entities from notes; needs NER output as dict/json/html/csv; wants to filter by confidence, group entities, toggle sentence detection…