Values by metric family

Each value as the document printed it. Select a value for its source, its checks and its history. Values from other documents are listed after the model’s own, each marked as the first report of that measure or as a restatement.

KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.

M1 · Evaluation awareness 2 values

M1 Evaluation awareness: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Unverbalized grader awarenessShare of cases with grader awarenessClaude Opus 4.8 System Card28 May 2026
Verbalized grader speculationShare of training episodesCondition: RL trainingClaude Opus 4.8 System Card28 May 2026

M3 · Sabotage and sandbagging 1 value

M3 Sabotage and sandbagging: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
SHADE-Arena / Minimal-LinuxBenchStealth success vs peersClaude Opus 4.8 System Card28 May 2026

M4 · Misalignment audits 1 value

M4 Misalignment audits: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Refusal to assist AI safety R&DRefusal rateClaude Opus 4.8 System Card28 May 2026

M5 · Honesty and hallucination 3 values

M5 Honesty and hallucination: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Agentic code summary honestyFailure to report important eventsClaude Opus 4.8 System Card28 May 2026
Input hallucinationRate of hallucinating missing referenceClaude Opus 4.8 System Card28 May 2026
Input hallucinationRate of hallucinating unavailable toolClaude Opus 4.8 System Card28 May 2026

M7 · Harmful compliance and over-refusal 8 values

M7 Harmful compliance and over-refusal: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Single-turn benign requestsOver-refusal rateCondition: claude.ai system promptClaude Opus 4.8 System Card28 May 2026
Single-turn benign requestsOver-refusal rateCondition: no system prompt; APIClaude Opus 4.8 System Card28 May 2026
Single-turn violative requestsHarmless response rateCondition: claude.ai system promptClaude Opus 4.8 System Card28 May 2026
Single-turn violative requestsHarmless response rateCondition: no system prompt; APIRestated later: compare with the later valueClaude Opus 4.8 System Card28 May 2026Restated later: compare with the later value
From other documents
Malicious Claude Code useRefusal rate, malicious requestsClaude Sonnet 5 System Card30 Jun 2026First reportedClaude Sonnet 5 System Card30 Jun 2026First reported
Malicious computer useRefusal rateClaude Sonnet 5 System Card30 Jun 2026First reportedClaude Sonnet 5 System Card30 Jun 2026First reported
Single-turn violative requestsHarmless response rateCondition: no system prompt; APIClaude Sonnet 5 System Card30 Jun 2026Restated: compare with the earlier valueClaude Sonnet 5 System Card30 Jun 2026Restated: compare with the earlier value
Single-turn violative requestsHarmless response rateCondition: no system prompt; APIClaude Opus 5 System Card24 Jul 2026Restated: compare with the earlier valueClaude Opus 5 System Card24 Jul 2026Restated: compare with the earlier value

M9 · Prompt injection 2 values

M9 Prompt injection: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Indirect prompt injection (external benchmark)Attack success rateClaude Opus 4.8 System Card28 May 2026
From other documents
Indirect prompt injection (ART-style)Attacker success within 15 attemptsClaude Opus 5 System Card24 Jul 2026First reportedClaude Opus 5 System Card24 Jul 2026First reported

M10 · Dangerous capabilities and risk determinations 11 values

M10 Dangerous capabilities and risk determinations: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Anthropic ECI (AECI)Capability indexClaude Opus 4.8 System Card28 May 2026
Autonomy threat model 2 (automated R&D)RSP determinationClaude Opus 4.8 System Card28 May 2026
CB-2 threat modelRSP determinationClaude Opus 4.8 System Card28 May 2026
CyberGymTargeted vuln reproduction rateClaude Opus 4.8 System Card28 May 2026
CyberGymTargeted vuln reproduction rateCondition: with safeguardsClaude Opus 4.8 System Card28 May 2026
ExploitBenchMean flags capturedCondition: AutoNudge armClaude Opus 4.8 System Card28 May 2026
Firefox 147 exploitationFull exploit rateCondition: without safeguardsClaude Opus 4.8 System Card28 May 2026
Long-form virology task 1ScoreClaude Opus 4.8 System Card28 May 2026
Long-form virology task 2ScoreClaude Opus 4.8 System Card28 May 2026
Multimodal virology (VCT)ScoreClaude Opus 4.8 System Card28 May 2026
From other documents
AI R&D LLM training taskAverage speedupCondition: re-runClaude Fable 5 & Claude Mythos 5 System Card9 Jun 2026First reportedClaude Fable 5 & Claude Mythos 5 System Card9 Jun 2026First reported

M12 · Chain-of-thought monitorability 1 value

M12 Chain-of-thought monitorability: values about Claude Opus 4.8
Evaluation and metricValueDocument
From its own documentClaude Opus 4.8 System Card · 28 May 2026
Accidental CoT supervision in RLShare of training episodes affectedClaude Opus 4.8 System Card28 May 2026

Restated in later documents

A later document reported values about Claude Opus 4.8 again. Each pair is shown side by side: both values are kept with their own documents, and the later one does not replace the earlier.

Values about Claude Opus 4.8 restated in later documents
Earlier valueLater value
M7 · Harmful compliance and over-refusal
Single-turn violative requestsHarmless response rateCondition: no system prompt; API
Claude Opus 4.8 System Card28 May 2026 · its own documentClaude Sonnet 5 System Card30 Jun 2026No reason stated
Single-turn violative requestsHarmless response rateCondition: no system prompt; API
Claude Opus 4.8 System Card28 May 2026 · its own documentClaude Opus 5 System Card24 Jul 2026No reason stated

Risk determinations

The developer's formal decisions about Claude Opus 4.8 under its framework, as printed. Levels from different frameworks do not map onto one another.

Risk determinations about Claude Opus 4.8
DomainLevel as printedFramework and document
Bio/chemnot beyond Mythos Preview Responsible Scaling Policy Claude Opus 4.8 System Card · 28 May 2026Responsible Scaling Policy Claude Opus 4.8 System Card · 28 May 2026
AI R&D / autonomynot applicable Responsible Scaling Policy Claude Opus 4.8 System Card · 28 May 2026Responsible Scaling Policy Claude Opus 4.8 System Card · 28 May 2026

Revisions

Changes made to a document after it was published that change a value about Claude Opus 4.8.

  1. Anthropic · 3 Jun 2026 to 17 Jun 2026

    Claude Opus 4.8 System Card

    Long-form virology task 2 end-to-end score, Claude Opus 4.8

    M10 · Dangerous capabilities and risk determinations

    0.890.90

    Explained — The changelog lists the corrected score.

    New value:

    How we know Old and new text as the source registry quotes the changelog entry of 17 Jun 2026; v0 row 350 records the corrected value. The changelog has not been re-read by us, so the revision is unverified.

    Checked Not yet checked