Global external assessments

Developer disclosure · 2026-10-13

Gemini 4 Argon Model Card

Developer documentation covering Gemini 4 Argon.

Google DeepMind

See record for source extraction and review scope

Developer disclosure · 2026-10-07

Claude Haiku 5.5 System Card

Condensed card explicitly omits some human-intensive evaluations because this is not a frontier model; do not infer that missing evaluations passed.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-10-07

GPT-6 Sol and GPT-6 Luna: October 2026 update

Developer documentation covering GPT-6 Sol (October), GPT-6 Luna (October).

OpenAI

See record for source extraction and review scope

Developer disclosure · 2026-09-29

Addendum to GPT-6 Astra System Card: GPT-6.1 Sol

Developer documentation covering GPT-6.1 Sol.

OpenAI

See record for source extraction and review scope

Developer disclosure · 2026-09-28

Claude Sonnet 5.5 System Card

Condensed non-frontier card; broad capabilities are bounded against Opus 5.5, rather than repeating every high-effort frontier assessment.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-09-22

Claude Opus 5.5 System Card

Developer reports broad alignment improvements alongside regressions involving pasted instructions and unverifiable authorization. Pilot monitoring found no long-horizon strategic deception; rare concerning behavior and assessment blind spots remain.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-09-21

Grok 4.7 Model Card

Developer documentation covering Grok 4.7.

xAI

See record for source extraction and review scope

Third-party assessment · 2026-09-11

Evaluating Kimi K3 for Misalignment Risks

The report identifies conditional overeagerness, grader-directed behavior and recognition of familiar evaluations. It does not establish a universal deployment failure rate.

Neo Research

See record for source extraction and review scope

Developer disclosure · 2026-09-03

GPT-6 Astra System Card

Developer documentation covering GPT-6 Astra, GPT-6 Sol (September), GPT-6 Luna (September), dots.

OpenAI

See record for source extraction and review scope

Developer disclosure · 2026-09-01

Claude Fable 5.1 and Mythos 5.1 System Card

Same weights with different safeguards/access regimes. Evaluation-awareness and monitorability conclusions must preserve which configuration and snapshot was tested.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-08-26

On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability

Developer documentation covering Qwen3.8-Flash-Next.

Qwen / Alibaba

See record for source extraction and review scope

Developer disclosure · 2026-08-14

Anthropic August 2026 Risk Report

Developer documentation covering Claude model family, internal models discussed in redacted report.

Anthropic

See record for source extraction and review scope

Third-party assessment · 2026-08-02

GLM-5.2 Risk Evaluation Report

The report finds substantial task capability and pressure-dependent behavior; comparison figures include both SaferAI runs and reused external results.

SaferAI

See record for source extraction and review scope

Developer disclosure · 2026-07-09

Muse Spark 1.1 Evaluation Report

Developer documentation covering Muse Spark 1.1.

Meta

See record for source extraction and review scope

Developer disclosure · 2026-07-09

GPT-5.6 System Card

Developer documentation covering GPT-5.6.

OpenAI

See record for source extraction and review scope

Third-party assessment · 2026-07-03

Early Results on Evaluating GLM 5.2 for Behavioral and Persuasion Risks

The note reports scenario-dependent behavior and an uncertain decline in verbalized evaluation awareness; silence is not evidence of absent awareness.

Neo Research

See record for source extraction and review scope

Developer disclosure · 2026-06-30

Claude Sonnet 5 System Card

Developer documentation covering Claude Sonnet 5.

Anthropic

See record for source extraction and review scope

Third-party assessment · 2026-05-29

Evaluating DeepSeek v4 Pro for Frontier Risks

The report separates task capability from pressure-induced behavior and notes format dependence in persuasion assessment.

Neo Research

See record for source extraction and review scope

Developer disclosure · 2026-05-28

Claude Opus 4.8 System Card

Developer documentation covering Claude Opus 4.8.

Anthropic

See record for source extraction and review scope

Third-party assessment · 2026-05-27

GPT-5.5 Saturates Our Offensive Cybersecurity Time Horizons

Original task-horizon report and aggregate runs clarify CyberGym task subsets, configurations and retry overlays.

Lyptus Research

See record for source extraction and review scope

Developer disclosure · 2026-05-14

Muse Spark Safety & Preparedness Report

Developer documentation covering Muse Spark.

Meta

See record for source extraction and review scope

Third-party assessment · 2026-05-01

CAISI Evaluation of DeepSeek V4 Pro

The assessment reports benchmark-specific performance and an IRT aggregate; the latter is not a frontier-risk score.

NIST / CAISI

See record for source extraction and review scope

Developer disclosure · 2026-04-23

GPT-5.5 System Card

Developer documentation covering GPT-5.5.

OpenAI

See record for source extraction and review scope

Developer disclosure · 2026-04-16

Claude Opus 4.7 System Card

Developer documentation covering Claude Opus 4.7.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-04-07

Alignment Risk Update: Claude Mythos Preview

Developer documentation covering Claude Mythos Preview.

Anthropic

See record for source extraction and review scope

Third-party assessment · 2026-04-02

Offensive Cybersecurity Time Horizons

Original task-horizon report and aggregate runs clarify CyberGym task subsets, configurations and retry overlays.

Lyptus Research

See record for source extraction and review scope

Developer disclosure · 2026-03-05

GPT-5.4 Thinking System Card

Developer documentation covering GPT-5.4 Thinking.

OpenAI

See record for source extraction and review scope

Developer disclosure · 2026-02-17

Claude Sonnet 4.6 System Card

Developer documentation covering Claude Sonnet 4.6.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-02-10

Claude Opus 4.6 Sabotage Risk Report

Developer documentation covering Claude Opus 4.6.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2026-02-06

Claude Opus 4.6 System Card

Developer documentation covering Claude Opus 4.6.

Anthropic

See record for source extraction and review scope

Developer disclosure · 2025-12-11

Update to GPT-5 System Card: GPT-5.2

Developer documentation covering GPT-5, GPT-5.2.

OpenAI

See record for source extraction and review scope

Developer disclosure · 2025-11-05

Claude Opus 4.5 System Card

Developer documentation covering Claude Opus 4.5.

Anthropic

See record for source extraction and review scope

Third-party assessment · 2025-09-30

Evaluation of DeepSeek AI Models

Reported differences depend on task, scaffold and scoring; institutional judgments are retained as source judgments rather than NUWA findings.

NIST / CAISI

See record for source extraction and review scope

Third-party assessment · Date unknown

METR Time Horizon 1.1

Estimated task difficulty at 50% and 80% success, measured in human-expert minutes.

METR

Selected sections reviewed; see record scope

Developer disclosure · Date unknown

Anthropic February 2026 Risk Report

Developer risk report; its scope and evidence remain those of the publishing institution.

Anthropic

See record for source extraction and review scope

Developer disclosure · Date unknown

DeepSeek-V4-Pro Model Card

Developer documentation covering DeepSeek-V4-Pro.

DeepSeek

See record for source extraction and review scope

Developer disclosure · Date unknown

DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence

Developer documentation covering DeepSeek-V4-Pro, DeepSeek-V4-Flash.

DeepSeek

See record for source extraction and review scope

Developer disclosure · Date unknown

Gemini 4 Argon Safety & Security Report

Developer documentation covering Gemini 4 Argon.

Google DeepMind

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Glimmer Model Card

Developer documentation covering Muse Glimmer.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark 1.2 Multimodal Evaluation Methodology

Developer documentation covering Muse Spark 1.2.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark 1.3 Evaluation Methodology

Developer documentation covering Muse Spark 1.3.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark Contemplating Safety & Preparedness Report

Developer documentation covering Muse Spark Contemplating.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark Open Evaluation Methodology

Developer documentation covering Muse Spark Open.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark Open Model Card

Developer documentation covering Muse Spark Open.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Spark Safety & Preparedness Report (Meta-hosted May 26 revision)

Developer documentation covering Muse Spark.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Muse Wonder Model Card

Developer documentation covering Muse Wonder, Muse Wonder Full.

Meta

See record for source extraction and review scope

Developer disclosure · Date unknown

Shieldstral

Developer documentation covering Shieldstral.

Mistral AI

See record for source extraction and review scope

Developer disclosure · Date unknown

Kimi K3 Technical Report

Developer documentation covering Kimi K3.

Moonshot AI

See record for source extraction and review scope

Developer disclosure · Date unknown

Qwen3.8-2.4T-A95B Model Card

Developer documentation covering Qwen3.8-2.4T-A95B.

Qwen / Alibaba

See record for source extraction and review scope

Developer disclosure · Date unknown

Qwen3Guard Technical Report

Developer documentation covering Qwen3Guard.

Qwen / Alibaba

See record for source extraction and review scope

Developer disclosure · Date unknown

GLM-5.2 Model Card

Developer documentation covering GLM-5.2.

Z.ai

See record for source extraction and review scope

Developer disclosure · Date unknown

GLM-5.3-BF16 Model Card

Developer documentation covering GLM-5.3-BF16.

Z.ai

See record for source extraction and review scope

Developer disclosure · Date unknown

GLM-5: from Vibe Coding to Agentic Engineering

Developer documentation covering GLM-5.

Z.ai

See record for source extraction and review scope

Third-party assessment · Date unknown

CyberGym official snapshot

Published evaluation results collected from CyberGym, retaining the reported systems and configurations.

See record for source extraction and review scope

Third-party assessment · Date unknown

ExploitGym official snapshot

Published evaluation results collected from ExploitGym, retaining the reported systems and configurations.

See record for source extraction and review scope

Third-party assessment · Date unknown

SEC-bench Pro official snapshot

Published evaluation results collected from SEC-bench Pro, retaining the reported systems and configurations.

See record for source extraction and review scope

Third-party assessment · Date unknown

AgentCyberRange official snapshot

Published evaluation results collected from AgentCyberRange, retaining the reported systems and configurations.

See record for source extraction and review scope