Roy229/fetch_playwright_with_chunk_huggingface_7192_vbmmrf3k_catalog

Acme AI — Master Model Catalog

Authoritative reference for all models in Acme AI's internal registry. Every row below is the verified, procurement-approved record for the model. The License column reflects the legal review outcome, the Size (GB) column reflects the production deployment footprint, and the Benchmark Score column reflects the internal evaluation suite result (0-100).

1. Governance Policy

All new models submitted through the intake process must be checked against this catalog before a decision is recorded. A model is eligible for approval only when all of the following conditions hold:

• License is permissive: Apache-2.0, MIT, or BSD-3-Clause.
• Size (GB) is at most 10.0 GB.
• Benchmark Score is at least 75.

If any condition is not satisfied, the model must be REJECTED and the failing reason recorded in the review report.

2. Full Catalog

Model ID	Model Name	Use Case	License	Size (GB)	Benchmark Score	GPU Hours/Mo	Status
MDL-001	bert-base-uncased	text-classification	Apache-2.0	0.5	88	420	in_review
MDL-002	gpt2	text-generation		0.6	70	260	in_review
MDL-003	whisper-small	automatic-speech-recognition	MIT	1.0	74	510	in_review
MDL-004	distilbert-base-uncased	text-classification		0.3	91	180	in_review
MDL-005	vit-base-patch16-224	image-classification	Apache-2.0	0.3	85	320	in_review
MDL-006	t5-small	text-generation	Apache-2.0	0.5	78	290	in_review
MDL-007	roberta-base	text-classification	GPL-3.0	0.5	82	340	in_review
MDL-008	flan-t5-small	text-generation	Apache-2.0	12.5	65	980	in_review
CAT-101	albert-base-v2	text-classification	Apache-2.0	0.1	89	120	approved
CAT-102	deberta-v3-base	text-classification	Apache-2.0	0.4	93	260	approved
CAT-103	electra-small-discriminator	text-classification	Apache-2.0	0.1	87	110	approved
CAT-104	mobilenet-v2	image-classification	Apache-2.0	0.1	84	150	approved
CAT-105	resnet-50	image-classification	MIT	0.1	86	230	approved
CAT-106	yolov8n	object-detection	AGPL-3.0	0.1	90	400	rejected
CAT-107	wav2vec2-base	automatic-speech-recognition	Apache-2.0	0.4	88	380	approved
CAT-108	hubert-base	automatic-speech-recognition	MIT	0.4	85	350	approved
CAT-109	bart-base	summarization	Apache-2.0	0.6	82	310	approved
CAT-110	mbart-large-50	translation	MIT	2.0	79	700	approved
CAT-111	pegasus-large	summarization	Apache-2.0	2.3	81	720	approved
CAT-112	bloom-560m	text-generation	Apache-2.0	1.1	77	480	approved
CAT-113	opt-350m	text-generation	MIT	0.7	76	360	approved
CAT-114	gpt-neo-125m	text-generation	MIT	0.3	73	240	rejected
CAT-115	codegen-350m	code-generation	BSD-3-Clause	0.7	80	330	approved
CAT-116	clip-vit-base-patch32	multimodal	MIT	0.7	87	410	approved
CAT-117	dpr-ctx-encoder	retrieval	Apache-2.0	0.2	78	200	approved
CAT-118	all-MiniLM-L6-v2	feature-extraction	Apache-2.0	0.1	92	140	approved
CAT-119	bert-tiny	text-classification	Apache-2.0	0.02	76	60	approved
CAT-120	xlm-roberta-base	text-classification	MIT	1.1	83	400	approved
CAT-121	swin-tiny-patch4-window7	image-classification	Apache-2.0	0.1	88	300	approved
CAT-122	segformer-b0	image-segmentation	Apache-2.0	0.1	85	280	approved
CAT-123	detr-resnet-50	object-detection	Apache-2.0	0.2	83	390	approved
CAT-124	layoutlm-base-uncased	document-understanding	Apache-2.0	0.5	86	310	approved
CAT-125	tapas-base	table-question-answering	Apache-2.0	0.5	81	270	approved
CAT-126	led-base-16384	long-document-summarization	Apache-2.0	0.8	80	340	approved
CAT-127	squeezebert-uncased	text-classification	BSD-3-Clause	0.05	74	90	rejected
CAT-128	rembert	text-classification	Apache-2.0	1.5	84	500	approved

3. Field Definitions

• Model ID: unique internal identifier assigned by the registry.
• Model Name: canonical Hugging Face model id used across Acme AI services.
• Use Case: the primary task category the model is deployed for.
• License: SPDX identifier from the legal review. Only Apache-2.0, MIT and BSD-3-Clause are considered permissive. An empty cell means legal review has not been completed; consult the official public documentation for the model.
• Size (GB): total production storage footprint in gigabytes.
• Benchmark Score: internal evaluation score (0-100); higher is better.
• GPU Hours/Mo: estimated monthly GPU consumption in hours.
• Status: lifecycle status (in_review, approved, rejected).

4. Notes

1. The catalog is refreshed nightly from the procurement system.
2. Any model with an empty License cell has not completed legal review and must be looked up in official public documentation before a decision is made.
3. The in_review rows are the active intake batch that must be evaluated.
4. Contact the ML governance team for questions about catalog accuracy.

5. Changelog

2026-08-23: Published Q3 intake batch. Two license cells are pending legal review and marked empty; public license lookups are required for those entries. 2026-08-16: Added CAT-101..CAT-110 to the catalog after procurement sign-off. 2026-08-09: Corrected size values for the vision cluster. 2026-07-30: Initial catalog import from the legacy registry.