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 agentmods add skills/blaizzy/mlx-vlm/hf-cache-modelsnpx 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.svg" alt="Measured on agentmods" height="20"></a>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 | $0.00060 | $0.00592 |
| Opus 5 | $0.00030 | $0.00296 |
| Sonnet 5 | $0.00012 | $0.00118 |
| Haiku 4.5 | $0.00006 | $0.00059 |
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 5d 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.
- whether the list came from the script or from `curl http://127.0.0.1:8080/v1/models` How it starts
The opening of the file, as written. The whole thing — 66 lines — stays where its author put it; the contents beside it link to each section on GitHub.
HF Cache Models
Use this workflow to list locally cached Hugging Face models that MLX-VLM can expose when Hugging Face cache discovery is explicitly enabled.
Supported Model Rule
Match the server /v1/models filter used with --model-discovery hf-cache:
- repo type is
model mainrevision exists in the cacheconfig.jsonexiststokenizer_config.jsonexists- either
model.safetensors.index.jsonexists or at least one*.safetensorsfile exists
This is a cache/file-presence check that mirrors the server's opt-in hf-cache discovery mode (mlx_vlm/server/app.py). It does not load the model, prove generation works, or affect the default served listing. Pass --check-arch to additionally require that mlx-vlm ships an architecture for the model_type — this narrows the list from a cache candidate to probably loadable (folder-name match; it does not resolve MODEL_REMAPPING aliases, so use it as a strong hint, not proof).
Script
Use the bundled script instead of rewriting cache-scanning logic:
uv run python skills/skills/hf-cache-models/scripts/list_supported_hf_cache_models.py
JSON output:
uv run python skills/skills/hf-cache-models/scripts/list_supported_hf_cache_models.py --json
Only models mlx-vlm can actually load (architecture present, not just files present):
uv run python skills/skills/hf-cache-models/scripts/list_supported_hf_cache_models.py --check-arch
Custom cache directory:
uv run python skills/skills/hf-cache-models/scripts/list_supported_hf_cache_models.py \
--cache-dir /path/to/huggingface/cache
Reporting
When reporting the result, include:
- cache directory used, if non-default
- number of supported models
- exact model IDs
- whether the list came from the script or from
curl http://127.0.0.1:8080/v1/models
For Hugging Face cache-discovery verification, start the server with
--model-discovery hf-cache and compare with:
curl http://127.0.0.1:8080/v1/models
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.
- 5d 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 (5,466 stars, last pushed yesterday), licensed MIT. It adds 60 tokens to every session and 592 once invoked, about $0.0003 per session on Opus 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…
annotating-variants
Annotates VCF variants and normalizes HGVS nomenclature with public, license-free annotators (Ensembl VEP REST, VEP/SnpEff/ANNOVAR offline) and links variants to gnomAD population frequencies and the clinical context OpenMed extracts. Use when the user wants to predict variant consequences, map HGVS to genomic…
auditing-part11-trails
Generates and verifies 21 CFR Part 11-style audit trails — who/what/when, electronic signatures, and tamper-evidence — for OpenMed pipelines in GxP and clinical-trial (GCP) settings. Use when the user runs OpenMed in a regulated/validated environment and needs an attributable, time-stamped, tamper-evident record of…
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…
building-patient-timelines
Assemble a chronological patient timeline from OpenMed-extracted clinical events, normalizing dates and resolving relative time expressions on-device. Use when the user wants to build a patient timeline, order events from clinical notes, reconstruct a longitudinal history, plot a course of illness, or turn…