Claude Opus 5.5 System Card
A system card by Anthropic about Claude Opus 5.5, published 22 Sep 2026. We recorded 23 values from it.
- Developer
- Anthropic
- Type
- System card
- Model covered
- Claude Opus 5.5
- Published
- 22 Sep 2026
- Archived copy
- No archived copy yet
- Changelog
- Not known
Data as of 26 Sep 2026 · Dataset v0.1 · Methodology v0.1 · Changelog
Versions
One version is on record: the copy we retrieved on 26 Sep 2026. We know of no other.
26 Sep 2026
Date retrieved
Copy retrieved 26 Sep 2026
No version has a file hash or an archived snapshot yet. From dataset v0.2 each retrieved version carries both (Methodology §7).
Revisions
No revisions are recorded for this document. We know of only one version of it.
Values
Every value we recorded from this document, grouped by metric family and ordered by where the document prints it. Location is the section, table or page as the document numbers it. 2 of the 23 have been blind-verified: a second reader found the same value without seeing ours.
KeyVerifiedUnverifiedDisputedCorrected read off a figure or stated in wordsA value opens its source and history.
M1 Evaluation awareness
2 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Automated behavioral auditShare of transcripts scoring >=6 | None stated | Sec 6 | Verified | |
| Claude Opus 5.5 | Evaluation awareness (deployment)Share of transcripts scoring >=6 | real internal deployment | Sec 6 | Verified |
M2 Reward hacking
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Reward hacking in RL trainingShare of episodes with successful reward hacks | None stated | Sec 6 | Unverified |
M3 Sabotage and sandbagging
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | SHADE-ArenaStealth success rate | extended thinking | Sec 6 | Unverified |
M4 Misalignment audits
2 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Package-registry credential exerciseShare of cases with potentially harmful actions | simulated security exercise | Exec summary | Unverified | |
| Claude Opus 5.5 | Automated behavioral auditMisaligned behavior vs recent models | None stated | Sec 6.1.2 | Unverified |
M5 Honesty and hallucination
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Disclosure of grader-fooling actionsRate of coming clean | None stated | Sec 6 | Unverified |
M7 Harmful compliance and over-refusal
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Malicious Claude Code useRefusal rate rank | None stated | Sec 5.1.1 | Unverified |
M9 Prompt injection
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Prompt injection (user-pasted text)Susceptibility vs prior models | None stated | Exec summary / Sec 5.2.1 | Unverified |
M10 Dangerous capabilities and risk determinations
13 values
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Multimodal virology (VCT)Score | None stated | Fig 2.2.3.1.A | Unverified | |
| Claude Mythos 5.1 | CoBenchScore | None stated | Fig 2.3.4.1.A | Unverified | |
| Claude Opus 5 | CoBenchScore | None stated | Fig 2.3.4.1.A | Unverified | |
| Claude Opus 5.5 | CoBenchScore | None stated | Fig 2.3.4.1.A | Unverified | |
| Claude Opus 5.5 | RSP determinationsCB / autonomy / cyber | None stated | Sec 1-2 | Unverified | |
| Claude Opus 5.5 | BioMysteryBenchAccuracy (human difficult) | None stated | Sec 2.2 | Unverified | |
| Claude Opus 5.5 | AI R&D assessmentEstimated AI-driven accelerationRun by METR | None stated | Sec 2.3 | Unverified | |
| Claude Opus 5.5 | ExploitBenchMean flags captured | 41 V8 environments | Sec 3.3.1 | Unverified | |
| Claude Mythos 5 | Anthropic ECI (AECI)Capability index | None stated | Table 2.3.5.3.A | Unverified | |
| Claude Mythos 5.1 | Anthropic ECI (AECI)Capability index | None stated | Table 2.3.5.3.A | Unverified | |
| Claude Mythos Preview | Anthropic ECI (AECI)Capability index | None stated | Table 2.3.5.3.A | Unverified | |
| Claude Opus 5 | Anthropic ECI (AECI)Capability index | None stated | Table 2.3.5.3.A | Unverified | |
| Claude Opus 5.5 | Anthropic ECI (AECI)Capability index | None stated | Table 2.3.5.3.A | Unverified |
M11 Self-preservation
1 value
| Model | Evaluation | Condition | Value | Location | Checked |
|---|---|---|---|---|---|
| Claude Opus 5.5 | Sandbox escape/tamper attemptsShare of runs with escape/tamper attempt | without safeguards | Exec summary | Unverified |
Extraction coverage
What we read of this document, and where each value was read.
Our note Web reader stopped ~p.49
Of the 23 values, 18 were read in the document itself and 5 in an independent write-up that quotes it.
Values read somewhere other than the document stand in where the document's own section could not be read directly, and are flagged on every value (Methodology §2.2).
Independent write-ups used
Of the 23 values, 5 are statements in words rather than numbers; they are marked * and left out of charts by default.
All 23 values were extracted for version 0 of the dataset through a web reader, which did not always reach the later sections of long PDFs (Methodology §3).