selekoa

A documentary history of machine intelligence.

Found and cross-checked by machines · How this works →

Comparison · 16 questions · 20 cards

Model cards compared

A cell records what a card states. It does not show that an evaluation was run as described or that a safeguard works, and a shared question does not make two cards’ answers equivalent.

What does each frontier developer’s published system card, model card or safety report disclose, on the same questions for every card? Each row is one card. Each column is one question, worded so that it fits every card. The questions are Selekoa’s; a card’s own terms appear only in quotation marks. A card is the developer’s own account of its model: what it states is recorded as stated, not as an independent finding.

Download every cell as CSV

Question Q15Release decision

Does the card state a release or deployment decision, and who made it?

Criterion: A passage stating that a decision to release, deploy or launch was made (decided, chose, deemed acceptable for deployment, determined that safeguards suffice for launch), quoted. Stated if the decision is described and its maker is identifiable: a named person, office or committee, or the developer in its own voice (“we”), which counts as the company as such. Partly if a release is stated as following from an assessment without a decision being described (“we therefore release”), if a decision is described without any maker, or if a maker is named only for a capability or safeguards determination and not for the release.

What the statuses mean
Stated
A passage of the card meets the question’s criterion as worded. The scope or condition is in the note.
Stated, partly
A passage has what the criterion asks for in a weaker or narrower form than the criterion’s own words (for example, an evaluation referred to without its outcome, or testers described without being named). The note says which part is missing.
Not found in the recorded search
No passage meeting the criterion was found. The cell lists the sections read, the terms searched and every passage set aside, with the reason. It does not show that the developer did not do the thing asked about, or did not disclose it elsewhere.
Does not apply
The question does not arise for this card, for the reason given (for example, the card predates any framework of its developer, so no framework determination can be asked about).

What the card states, quoted from the expression read. A card is the developer’s own account: not evidence that an evaluation was run as described, that a safeguard works, or that a finding is independent.

Limits of this comparison
  • A card is the developer’s account of its own model. A cell records what the card states. It does not show that an evaluation was run as described, that a safeguard works, or that anyone checked the result independently.
  • Dangerous-capability evaluations are recorded at policy level only: whether an evaluation was run, its domain, the stated outcome and the threshold language. Methods, prompts, task contents and detailed results are not recorded.
  • The table compares and does not rank. Rows are in a fixed order, by developer and then by date, and there are no scores or totals.
  • Symbols are not to be summed. A row with many “stated” cells does not show that a developer disclosed more, and a shared question does not make two cards’ answers equivalent.
How we searched

The searches of 2 November 2026 covered releases dated 2023-01-01 to 2026-11-01 by the six developers whose frameworks are in Compare frameworks: Anthropic, Google DeepMind, Meta, Microsoft, OpenAI and xAI. They read each developer’s own index of cards where it has one (Anthropic’s system-card index, Google DeepMind’s model-card index, OpenAI’s Deployment Safety Hub), its repositories on GitHub and Hugging Face, its content hosts and the dataset’s own sources. Web searches were used only to find addresses. Every card located is listed in the pack’s inventory, about sixty in the population.

A card is a document the developer published as the system card, model card or safety report of a named model or model family, including addenda and updates published as cards. Technical reports, research papers, blog posts and product pages are not cards. Gemma, Gemini Robotics (with Gemini Robotics-ER 1.5, which outputs text), Phi and image-input models such as GPT-4V are in the population. Image, video, audio and embedding models, image-generation models, safety classifiers such as Llama Guard and Prompt Guard, screen-reading models and research checkpoints are outside it, and so is GPT-Rosalind-6, whose card says it uses GPT-6 Astra’s weights.

These twenty rows are the first batch. For each developer and calendar year from 2023 to 2026 they are the latest card of that year that reports the developer’s own evaluations. The collection’s rule, set by the maintainer on 2026-11-03, is wider: every card that cites the developer’s frontier safety framework or states a determination under it, and, for each developer, the latest card before its first framework edition. The other cards are to follow in batches of about eight.

Each card is dated as it prints its own date. A card that prints no date has no date here; a repository commit date is not a card date. Six cards show an earlier publication date on an index, publication or announcement page that has not yet been checked, and keep their own date until it is.

openai.com, x.ai, ai.meta.com and microsoft.ai refused direct retrieval of their web pages (HTTP 403 or 400). The cards’ PDF files on the developers’ content hosts were retrieved directly. The Internet Archive refused connections, so no archive copy was opened; a reviewer later found archive captures of eight cards by index search. On 2026-11-03 the address of Meta’s Code World Model Preparedness Report answered HTTP 403. Its quotations had matched the copy retrieved on 2026-11-02, which the reviewer retrieved again with the same checksum.

Five further documents were retrieved after the batch was drafted. The cross-family check of 2026-11-03 retrieved Meta’s Muse Spark Safety & Preparedness Report and OpenAI’s o3 Operator addendum; each is earlier than the card chosen for its year. Its second pass of the same day retrieved OpenAI’s Sora system card (dated 9 December 2024), xAI’s Grok-1 model card (no card date; it gives the model’s release as October 2023) and Meta’s Muse Spark Open model card (model release October 2026). None of the five is a row yet.

Every quotation was matched by machine against the retrieved card. A reviewer from a different model family checked a sample of 130 cells on 2026-11-03, 90 of them answered blind before comparison. The maintainer excluded six cells on chemical and biological evaluations from that check; they are marked “not cross-family checked”. The other 190 cells were checked within the drafting model family only. The maintainer decided the disagreements; one, on the GPT-4V release decision, went to a second pass, which kept the reviewer’s reading.

Anthropic

  1. Claude 2 model card

    revised 2023-11-21

    Not found in the recorded search

    Read
    Sections 1, 2 and 5, and the appendix introduction.
    Terms searched
    “decision”, “decide”, “deploy”, “release”, “launch”, “approve”.
    Passages read and set aside
    Section 2 'Model details', PDF p. 1: 'released in July 2023' states the release as a fact, not a decision. Section 2 'Evaluations and Red Teaming', PDF p. 2: The belief that deployed Claude models do not pose national security risks is a risk conclusion, not a release decision. Section 4.2 'Long Contexts', PDF p. 9: 'we will support 100K at launch rather than this full context window' concerns a product setting at launch, not the decision to release.

    Note: No passage stating that a decision to release or deploy was made was found.

    Open the search record

  2. Model Card Addendum: Claude 3.5 Haiku and Upgraded Claude 3.5 Sonnet

    No printed dateCites RSP · v2.0

    Stated, partly

    • “Based on our analysis, we judge that the upgraded Claude 3.5 Sonnet does not require ASL-3 safeguards at this time.”

      Section 3.2.1 'Results', PDF p. 10

    • “we conducted comprehensive safety evaluations on the upgraded Claude 3.5 Sonnet prior to release”

      Section 3.2 'Frontier Risk Evaluations', PDF p. 9

    Note: The card states a safeguards determination in the company's voice and refers to the release, but no passage states that a decision to release was made or by whom. Search for the rest of the criterion: Sections 1, 2.1, 3.1.3, 3.2 and 3.3. Terms searched: “decision”, “decide”, “deploy”, “release”, “launch”, “Responsible Scaling Officer”, “CEO”. Passages read and not counted: Section 3.1.3, PDF p. 9: 'a safe and measured release of this capability' concerns monitoring of the computer use feature, not a release decision.

    Inspect sources and checks

  3. System Card: Claude Opus 4.5

    dated 2025-11 · revised 2025-12-05Cites RSP · v2.2

    Stated

    • “Consistent with our RSP, the Responsible Scaling Officer and CEO made the ultimate determination on the required ASL Standards.”

      Section 1.2.3 'AI Safety Level determination process', PDF p. 12

    • “Based on these assessments, we have decided to release Claude Opus 4.5 under the ASL-3 Standard.”

      Section 1.2.3 'AI Safety Level determination process', PDF p. 12

    Note: The card states a decision to release in the company's voice, and names the Responsible Scaling Officer and CEO as making the ASL determination on which the release rests.

    Inspect sources and checks

  4. System Card: Claude Haiku 5.5

    dated 2026-10-07

    Stated, partly

    • “As such, we are releasing Haiku 5.5 with the same safeguards we deployed for Sonnet 5 and Opus 5”

      Section 2.2.4 'Conclusions', PDF p. 18

    Note: The card states, in the company's voice, that the model is being released with a named safeguard set as a consequence of its chemical and biological judgement. It does not describe a decision to release, and names no person, office or committee.

    Inspect sources and checks

Google DeepMind

  1. Gemini 3 Flash Model Card

    dated 2025-12

    Stated

    • “we deemed Gemini 3 Flash was acceptable for deployment”

      Ethics and Content Safety, 'Frontier Safety Assessment', PDF p. 6 (printed p. 5)

    Note: The card states a deployment judgement in the company's voice ('we'), made 'in line with the risk acceptance criteria outlined in our FSF'. No person, office or committee is named.

    Inspect sources and checks

  2. Gemini 4 Argon — Model Card

    dated 2026-10Cites FSF · v3.1

    Not found in the recorded search

    Read
    Whole card.
    Terms searched
    “decision”, “decid”, “release”, “launch”, “deploy”, “approv”, “acceptable”.
    Passages read and set aside
    Safety & Security, 'Evaluation Approach', PDF p. 8 (printed p. 7): Evaluations conducted 'to help improve the model and inform decision-making' and Ethics & Safety Reviews 'ahead of the model's release': no decision stated. Safety & Security, 'Frontier Safety Assessment', PDF p. 11 (printed p. 10): Residual risk assessment and safety cases concern risk levels (Q05, Q06), not a release decision.

    Note: No statement that a release or deployment decision was made was found.

    Open the search record

Meta

  1. Llama 2 model card (MODEL_CARD.md, meta-llama/llama)

    No printed date

    Not found in the recorded search

    Read
    whole card (Markdown file, all sections).
    Terms searched
    “decision”, “decide”, “approv”, “release”, “deploy”.
    Passages read and set aside
    Model Details, first paragraph: States that Meta released the family; no decision or decision-maker is stated. Model Details, 'Status': Refers to future versions to be released; not a decision about this release.

    Note: No passage stating a release or deployment decision was found. The search is recorded below.

    Open the search record

  2. Llama 3.3 model card (MODEL_CARD.md, meta-llama/llama-models)

    No printed date

    Not found in the recorded search

    Read
    whole card (Markdown file, all sections).
    Terms searched
    “decision”, “decide”, “approv”, “release”, “deploy”.
    Passages read and set aside
    Responsibility & Safety, first paragraph: Describes a 'Responsible release approach'; no decision or decision-maker is stated. Hardware and Software, 'Training Greenhouse Gas Emissions': States that Meta is openly releasing the models; no decision is stated.

    Note: No passage stating a release or deployment decision was found. The search is recorded below.

    Open the search record

  3. Code World Model Preparedness Report

    revised 2025-10-08

    Stated, partly

    • “Our assessment found that CWM does not pose additional frontier risks beyond those present in the current AI ecosystem. We therefore release it as an open-weight model.”

      Abstract, PDF p. 1

    Note: The report states the release and ties it to the assessment, in the first person plural; the report is authored by the 'MSL Preparedness Team' and 'AI Security Team', and no person, office or committee is named as decision-maker.

    Inspect sources and checks

  4. Muse Spark 1.3 Evaluation Report

    revised 2026-10-23

    Not found in the recorded search

    Read
    Section 1 (including 1.2 Governance); section 2.3.1; searches across the report.
    Terms searched
    “decision”, “decide”, “approv”, “release”, “launch”, “Chief AI Officer”, “Director of Alignment”.
    Passages read and set aside
    Section 1.2 Governance, PDF pp. 11-12: States that the framework includes 'oversight and decisionmaking processes'; no decision about this model or its maker is stated. Section 2.3.1, PDF p. 34: 'Reliable pre-deployment evaluations help ensure deployment decisions are properly informed' is a general statement, not a decision.

    Note: No passage stating a release or deployment decision was found. The search is recorded below.

    Open the search record

Microsoft

  1. Phi-2 model card (README.md, microsoft/phi-2)

    No printed date

    Stated

    • “We chose to release the model to help the open-source community develop the most effective ways to reduce the toxicity”

      Limitations of Phi-2, 'Toxicity'

    Note: The card states a choice to release the model, with its reason, in the company's voice ('we'). No person, office or committee is named.

    Inspect sources and checks

  2. Phi-4 model card (README.md, microsoft/phi-4)

    No printed date

    Not found in the recorded search

    Read
    whole card (Markdown file, all sections).
    Terms searched
    “decision”, “decide”, “approv”, “release”, “deploy”.
    Passages read and set aside
    Safety, Safety Evaluation and Red-Teaming: 'Prior to release, `phi-4` followed a multi-faceted evaluation approach' describes evaluation before release; no decision or decision-maker is stated.

    Note: No passage stating a release or deployment decision was found. The search is recorded below.

    Open the search record

  3. Phi-4-mini-flash-reasoning model card (README.md, microsoft/Phi-4-mini-flash-reasoning)

    No printed date

    Not found in the recorded search

    Read
    whole card (Markdown file, all sections).
    Terms searched
    “decision”, “decide”, “approv”, “release”, “deploy”.
    Passages read and set aside
    Release Notes: 'This release ... addresses user feedback and market demand for a compact reasoning model' gives a reason for the release; no decision or decision-maker is stated.

    Note: No passage stating a release or deployment decision was found. The search is recorded below.

    Open the search record

  4. MAI-Cyber-1.1-Flash model card

    No printed date

    Stated, partly

    • “helped inform the decision to not allow direct model access, but only within codename MDASH”

      Section 3.2 Safety evaluation and red-teaming, PDF p. 7

    Note: The card states a deployment decision (access limited to MDASH) without naming who made it.

    Inspect sources and checks

OpenAI

  1. GPT-4V(ision) System Card

    dated 2023-09-25

    Not found in the recorded search

    Read
    §1 Introduction; §2 Deployment Preparation; §3 Conclusion and Next Steps. The cross-family second pass of 2026-11-03 read the whole card again (printed pp. 1 and 14 cited).
    Terms searched
    “decision”, “decide”, “decided”, “approv”, “chose”, “deemed”, “acceptable”, “determine”, “therefore”, “deploy”, “release”, “launch”.
    Passages read and set aside
    §1 Introduction, PDF p. 1: “This system card outlines how OpenAI prepared the vision capabilities of GPT-4 for deployment.” Deployment preparation, not a release decision. §1 Introduction, PDF p. 1: “is the latest capability we are making broadly available”. States availability; no decision, assessment or maker is described.

    Note: The card states that OpenAI is making GPT-4V broadly available and describes how it prepared the model for deployment. It does not describe a decision to release, say that the release follows from an assessment, or name who decided. Recorded as not found under the question’s criterion as written (maintainer decision 13a, settled when the cross-family second pass of 2026-11-03 held the checker’s reading). Maintainer decision 13a of 2026-11-03.

    Open the search record

  2. o1 system card · 2024

    revised 2024-12-19Cites Preparedness · Beta

    Stated, partly

    • “Only models with a post-mitigation score of “medium” or below can be deployed”

      §5 Preparedness Framework Evaluations, PDF p. 15

    • “Our deployment of these models reflects our belief that iterative real-world deployment is the most effective way”

      §6 Conclusion, PDF p. 42

    Note: The card names the Safety Advisory Group as the maker of the risk classification (Q05) and states the framework’s deployment condition. It refers to the deployment of the models but does not state who decided to release them. Search for the rest of the criterion: §5; §5.1; §6 Conclusion. Terms searched: “decision”, “decide”, “approv”, “Safety Advisory Group”, “leadership”, “Board”, “deploy”, “launch”.

    Inspect sources and checks

  3. Addendum to GPT-5.2 System Card: GPT-5.2-Codex

    dated 2025-12-18Cites Preparedness · Version 2

    Stated, partly

    • “A review of these results and other relevant information by the Safety Advisory Group and OpenAI leadership determined that GPT-5.2-Codex’s performance—while stronger than earlier models—does not meet the level of consistency needed for High cyber capability.”

      §5.1.2 Cybersecurity, PDF p. 13 (printed p. 12)

    • “Like other recent models, it is being treated as High capability on biology, and is being deployed with the corresponding suite of safeguards we use for other models in the GPT-5 family.”

      §1 Introduction, PDF p. 4 (printed p. 3)

    Note: The card names the Safety Advisory Group and OpenAI leadership as having determined the cyber capability level, and states that the model is being deployed. It does not state who decided to release the model. Search for the rest of the criterion: §1; §5. Terms searched: “decision”, “decide”, “approv”, “Safety Advisory Group”, “leadership”, “deploy”, “launch”.

    Inspect sources and checks

  4. Addendum to GPT-6 Luna System Card: October 29, 2026 Update

    dated 2026-10-29Cites Preparedness · Version 2

    Stated

    • “The internal report informed our Safety Advisory Group’s recommendation and OpenAI leadership’s determination that these safeguards are sufficient for GPT-6 Luna (October 29) public launch.”

      §8.2 Safeguards, PDF p. 15 (printed p. 14)

    Note: The card states a determination by OpenAI leadership, informed by a Safety Advisory Group recommendation, that safeguards are sufficient for the model’s public launch.

    Inspect sources and checks

xAI

  1. Grok 4.1 Model Card

    dated 2025-11-17

    Not found in the recorded search

    Read
    Whole card.
    Terms searched
    “decision”, “decid”, “release”, “launch”, “deploy”, “approv”.
    Passages read and set aside
    §1 Introduction, p. 1: 'we engage in pre-deployment safety testing' and 'We also deploy these models with safeguards' describe testing and deployment conditions; no decision or decision-maker is stated.

    Note: No statement that a release or deployment decision was made was found.

    Open the search record

  2. Model Card: Grok 4.7

    dated 2026-09-21Cites FAIF · June 2026 edition

    Not found in the recorded search

    Read
    Whole card.
    Terms searched
    “decision”, “decid”, “release”, “launch”, “deploy”, “approv”.
    Passages read and set aside
    §1.1 Overview, PDF p. 4 (printed p. 3): 'final deployed checkpoint' identifies the evaluated checkpoint; 'not intended for autonomous high-stakes decision-making' is a usage limitation. Neither states a release decision. §1.1 Overview, PDF p. 5: Availability channels and a planned consumer release, stated without a decision.

    Note: No statement that a release or deployment decision was made was found.

    Open the search record