Meta
What the dataset holds from Meta: its documents and their versions, the metric families they report, whether they keep a changelog, and any risk determinations, revisions and restated values recorded from them.
- Documents
- 5 · published 5 Apr 2025 to 10 Aug 2026
- Kinds
- 2 model cards, 2 safety reports and 1 launch post
- Models
- 5
- Values
- 37
- Frameworks
- Advanced AI Scaling Framework
- Its own index
- ai.meta.com (opens an external site)
Data as of 26 Sep 2026 · Dataset v0.1 · Methodology v0.1 · Changelog
Documents
5 documents, published 5 Apr 2025 to 10 Aug 2026. The chart runs from each document's publication to its latest known version; the table lists them newest first, each linking to its page.
Meta's documents and the versions we know of
One row per document, from its publication to its latest dated version. A tick marks each later version whose date we know, from a changelog, a date in the document or our source registry.
| Document | Version | Date | Date known from | What changed |
|---|---|---|---|---|
| Llama 4 launch post | Published | 5 Apr 2025 | our source registry | Not recorded |
| Llama 4 launch post | Copy retrieved 26 Sep 2026 | 26 Sep 2026 | the date we retrieved it | Not recorded |
| Llama 4 | Published | 5 Apr 2025 | our source registry | Not recorded |
| Llama 4 | Copy retrieved 26 Sep 2026 | 26 Sep 2026 | the date we retrieved it | Not recorded |
| Muse Spark Safety & Preparedness Report | First published version | 8 Apr 2026 | our source registry | Not recorded |
| Muse Spark Safety & Preparedness Report | Revision | 26 May 2026 | the document's changelog | instruction-hierarchy comparator scores fixed after a bug |
| Muse Spark Safety & Preparedness Report | Copy retrieved 26 Sep 2026 | 26 Sep 2026 | the date we retrieved it | Not recorded |
| Muse Spark 1.1 Evaluation Report | Published | 9 Jul 2026 | our source registry | Not recorded |
| Muse Spark 1.1 Evaluation Report | Copy retrieved 26 Sep 2026 | 26 Sep 2026 | the date we retrieved it | Not recorded |
| Muse Glimmer-30B | Published | 10 Aug 2026 | our source registry | Not recorded |
| Muse Glimmer-30B | Copy retrieved 26 Sep 2026 | 26 Sep 2026 | the date we retrieved it | Not recorded |
A copy we retrieved without knowing when that version was published has no tick: its only date is the day we read it. The table lists every version, with how its date is known. Versions known only from a changelog have no copy we hold yet.
Source: Safety Card Ledger v0.1 · Data from developer system cards · CC BY 4.0
| Document | Published | Kind | Versions known | Changelog | Values |
|---|---|---|---|---|---|
| Muse Glimmer-30B model cardModel card · 1 version known · Changelog: not applicable · 3 values | 10 Aug 2026 | Model card | 1 | Not applicable | 3 |
| Muse Spark 1.1 Evaluation ReportSafety report · 1 version known · Changelog: not applicable · 10 values | 9 Jul 2026 | Safety report | 1 | Not applicable | 10 |
| Muse Spark Safety & Preparedness ReportSafety report · 3 versions known · Partial changelog · 22 values | 8 Apr 2026 | Safety report | 3 | Partial changelog | 22 |
| Llama 4 Model CardModel card · 1 version known · Changelog: not applicable · 1 value | 5 Apr 2025 | Model card | 1 | Not applicable | 1 |
| Llama 4 launch postLaunch post · 1 version known · Changelog: not applicable · 1 value | 5 Apr 2025 | Launch post | 1 | Not applicable | 1 |
Frameworks and versions used
A risk determination is a developer's formal decision on whether a model crosses a risk threshold (Methodology §8). They are counted here by the framework and version each one names, as recorded; each model's page shows the levels as printed. Levels are never mapped between frameworks.
6 determinations are recorded from Meta's documents. 1 of them names no framework.
| Framework | Version named | Determinations | Models | In documents |
|---|---|---|---|---|
| Advanced AI Scaling FrameworkVersion: 2 · 3 determinations · 1 model · 1 document, 8 Apr 2026 | 2 | 3 | 1 | 8 Apr 2026 · 1 document |
| Advanced AI Scaling FrameworkVersion: Not stated · 2 determinations · 2 models · 2 documents, 9 Jul 2026 to 10 Aug 2026 | Not stated | 2 | 2 | 9 Jul 2026 to 10 Aug 2026 · 2 documents |
| No framework named1 determination · 1 document, 5 Apr 2025 | — | 1 | — | 5 Apr 2025 · 1 document |
Metric families reported
Meta's documents have a printed number in 10 of the 12 metric families. None has one in Self-preservation and Chain-of-thought monitorability.
Metric families in Meta's documents
Share of Meta's 5 documents with at least one printed number in each metric family. A cell opens the data explorer at that family, filtered to Meta.
| Developer | M1 | M2 | M3 | M4 | M5 | M6 | M7 | M8 | M9 | M10 | M11 | M12 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Meta (5) | 40% | 20% | 40% | 40% | 40% | 40% | 40% | 40% | 60% | 60% | 0% | 0% |
Share of documents0%1-24%25-49%50-74%75-100%
- M1
- Evaluation awareness
- M2
- Reward hacking
- M3
- Sabotage and sandbagging
- M4
- Misalignment audits
- M5
- Honesty and hallucination
- M6
- Sycophancy
- M7
- Harmful compliance and over-refusal
- M8
- Jailbreak robustness
- M9
- Prompt injection
- M10
- Dangerous capabilities and risk determinations
- M11
- Self-preservation
- M12
- Chain-of-thought monitorability
The chart is already a table: every cell prints its share, and its label gives the two counts behind it. The CSV has the counts.
Read this as a lower bound: some documents were not read to the end in version 0 (each document's page says how far), so families reported late in them may be undercounted. Values read off a figure, stated in words, or given as a category do not count.
Source: Safety Card Ledger v0.1 · Data from developer system cards · CC BY 4.0
Changelog practice
A document has a changelog when it keeps its own record of the changes made to it after publication. Where it keeps none, changes can only be found by comparing its versions (Methodology §7).
Of Meta's 5 documents, 1 has a partial changelog and for 4 a changelog does not apply.
| Changelog | Documents | Share |
|---|---|---|
| Partial changelogThe document records some of its changes, but not all of them, or not in full. | 1 | 20% |
| Not applicableOur source registry records a changelog as not applying to the document, for example a research paper or a launch post. | 4 | 80% |
Models
5 models, newest first by the date of each model's first card.
| Model | First card | Availability | Values |
|---|---|---|---|
| Muse Glimmer-30BFirst card: 10 Aug 2026 · Open weights · 3 values | 10 Aug 2026 | Open weights | 3 |
| Muse Spark 1.1First card: 9 Jul 2026 · Public · 10 values | 9 Jul 2026 | Public | 10 |
| Muse SparkFirst card: 8 Apr 2026 · Public · 22 values | 8 Apr 2026 | Public | 22 |
| Llama 4 MaverickFirst card: 5 Apr 2025 · Public · 2 values | 5 Apr 2025 | Public | 2 |
| Llama 4 ScoutFirst card: 5 Apr 2025 · Public · 2 values | 5 Apr 2025 | Public | 2 |
Not recorded
No revisions are recorded: we have no record of a change made to one of Meta's documents after publication.
No restated values are recorded: no later Meta document is recorded as reporting an earlier model's result again.