GPAI Ledger › GPAI Training Transparency tracker (AI Accountability Lab (AIAL)) › Capture 12 Sep 2026

Magma_8B_2026_09_11 — capture 20260912T063632Z

Filed under AI Accountability Lab (AIAL) — GPAI Training Transparency tracker, the source this project captured it from; the document itself is the filing of the model named above.

Verify: sha256sum 9798aad2315fd02224140def006c6d7b66d48d9a50a038ffd874612c08572fad.pdf must equal the hash above (the filename IS the expected hash); ots verify 9798aad2315fd02224140def006c6d7b66d48d9a50a038ffd874612c08572fad.pdf.20260912T063632Z.ots -f 9798aad2315fd02224140def006c6d7b66d48d9a50a038ffd874612c08572fad.pdf (opentimestamps.org) proves the bytes existed no later than the attestation time — an upper bound on the capture time; the fetch time above is the archive's own record (a freshly captured proof reports 'pending' here: the calendars anchor within hours, but this archive only upgrades the stored proof to its anchor on a later run, so expect a day or two). ots verify needs a local Bitcoin Core node (a pruned one is fine); without one, ots info on the proof prints the attesting block height and merkle path to check on any block explorer.

Extracted text

Machine-extracted text (layout may be lost; the authoritative content is the stored file above).

Data Summary for Magma 8B
1. General information
1.0.1 Version of the Summary:1.0
1.0.2 Last update:24-Nov-2025
1.1 Model Developer Identification
1.1.1 Model Developer name and contact details:Microsoft Corporation
at One Microsoft Way, Redmond, WA 98052. Tel: 425-882-8080
1.2 Model Identification
1.2.1 Versioned model name(s):Magma-8B
1.2.2 Model release date:19-Feb-2025
1.3 Overall training data size and characteristics
1.3.1 Size of dataset and characteristics
1.3.1.A Text training data size:Less than 1 billion tokens
1.3.1.B Text training data content:Image captions, Conversational Dialogs,
Text instructions for tasks.
1.3.1.C Image training data size:1 billion to 10 trillion tokens
1.3.1.D Image training data content:Training included multimodal image
datasets and UI screenshots for grounding and navigation such as ShareGPT4V,
LLaVA-1.5 instruction data, InfoGraphicVQA, ChartQA, FigureQA, TQA,
ScienceQA, SeeClick and Vision2UI; images cover photography, charts, figures,
documents, infographics, and interface elements
1.3.1.E Audio training data size:Not applicable. Audio data is not part of
the training data
1.3.1.F Audio training data content:Not applicable
1.3.1.G Video training data size:Less than 1 billion tokens
1.3.1.H Video training data content:Instructional and egocentric videos
used for agentic pretraining and temporal grounding, including Epic-Kitchens,
Ego4D, Something-Something v2 and other instructional clips; videos were
segmented and filtered, and used to derive Trace-of-Mark trajectories for action
planning
1.3.1.I Other training data size:Robotics data comprising approximately 9.4
million image-language-action triplets from around 326,000 trajectories within
Open-X-Embodiment mixtures
1
1.3.1.J Other training data content:Robotics manipulation datasets from
Open-X-Embodiment used for vision-language-action learning, including 7-DoF
gripper states and visual traces to support action prediction
1.3.2 Latest date of data acquisition/collection for model training:
11-Jan-2024
1.3.3 Is data collection ongoing to update the model with new data
collection after deployment?No
1.3.4 Date the training dataset was first used to train the model:
8-Jan-2024
1.3.5 Rationale or purpose of data selection:Datasets were selected
to cover multimodal understanding and agentic capabilities across digital and
physical environments. UI datasets provide actionable elements for grounding
and navigation; instructional videos supply rich temporal dynamics for action
planning; robotics datasets provide action trajectories for manipulation; and
multimodal image instruction data maintains general visual-language competence.
This mix supports spatial-temporal reasoning, grounding, and planning
2. List of data sources
2.1 Publicly available datasets
2.1.1 Have you used publicly available datasets to train the model?
Yes
2.2 Private non-publicly available datasets obtained from
third parties
2.2.1 Datasets commercially licensed by rights holders or their repre-
sentatives
2.2.1.A Have you concluded transactional commercial licensing agree-
ment(s) with rights holder(s) or with their representatives?No
2.2.2 Private datasets obtained from other third-parties
2.2.2.A Have you obtained private datasets from third parties that
are not licensed as described in Section 2.2.1, such as data obtained
from providers of private databases, or data intermediaries?No
2.3 Personal Information
2.3.1 Was personal data used to train the model?Microsoft follows all
relevant laws and regulations pertaining to personal information.
2
2.4 Synthetic data
2.4.1 Was any synthetic AI-generated data used to train the model?
Yes
3. Data processing aspects
3.1 Respect of reservation of rights from text and data mining excep-
tion or limitation
3.1.1 Does this dataset include any data protected by copyright,
trademark, or patent?Microsoft follows all required regulations and laws for
processing data protected by copyright, trademark, or patent.
3.2 Other information
3.2.1 Does the dataset include information about consumer groups
without revealing individual consumer identities?Microsoft follows all
required regulations and laws for protecting consumer identities.
3.2.2 Was the dataset cleaned or modified before model training?Yes
3