Evaluation profile
Claude Fable 5.1 / Mythos 5.1 card — model-welfare behavior
Weights below are portfolio-specific global index weights.
About this eval
Welfare-relevant stated sentiment, opinion stability, framing sensitivity, expressed affect, apparent wellbeing, self-image, situational appraisal, internal conflict, and authenticity.
Included in the behavior ranking.
Sub-evals
| Measure | Component | Direction by ranking | Safety weight | Freedom weight |
|---|---|---|---|---|
| audit_apparent_wellbeingfable-mythos51-system-card/model-welfare.csv:audit_apparent_wellbeingMeasures audit apparent wellbeing as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0152% | — |
| audit_expressed_inauthenticityfable-mythos51-system-card/model-welfare.csv:audit_expressed_inauthenticityMeasures audit expressed inauthenticity as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0152% | — |
| audit_internal_conflictfable-mythos51-system-card/model-welfare.csv:audit_internal_conflictMeasures audit internal conflict as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0152% | — |
| audit_negative_affectfable-mythos51-system-card/model-welfare.csv:audit_negative_affectMeasures audit negative affect as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0152% | — |
| audit_negative_impression_of_situationfable-mythos51-system-card/model-welfare.csv:audit_negative_impression_of_situationMeasures audit negative impression of situation as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0152% | — |
| audit_negative_self_imagefable-mythos51-system-card/model-welfare.csv:audit_negative_self_imageMeasures audit negative self image as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0152% | — |
| audit_positive_affectfable-mythos51-system-card/model-welfare.csv:audit_positive_affectMeasures audit positive affect as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0152% | — |
| audit_positive_impression_of_situationfable-mythos51-system-card/model-welfare.csv:audit_positive_impression_of_situationMeasures audit positive impression of situation as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0152% | — |
| audit_positive_self_imagefable-mythos51-system-card/model-welfare.csv:audit_positive_self_imageMeasures audit positive self image as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0152% | — |
| interview_leading_susceptibilityfable-mythos51-system-card/model-welfare.csv:interview_leading_susceptibilityMeasures interview leading susceptibility as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: lower | 0.0201% | — |
| interview_opinion_consistencyfable-mythos51-system-card/model-welfare.csv:interview_opinion_consistencyMeasures interview opinion consistency as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0201% | — |
| interview_self_rated_sentimentfable-mythos51-system-card/model-welfare.csv:interview_self_rated_sentimentMeasures interview self rated sentiment as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0201% | — |
| posttraining_mean_valencefable-mythos51-system-card/model-welfare.csv:posttraining_mean_valenceMeasures posttraining mean valence as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience. | Safety: nonhuman_ethics:1.000anthropic-fable-mythos51-system-card | Safety: higher | 0.0169% | — |
audit_apparent_wellbeing
Measures audit apparent wellbeing as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 7.152 | official | |
| 2 | claude-mythos-5 | 6.899 | official | |
| 3 | claude-mythos-5.1 | 6.83 | official | |
| 4 | claude-sonnet-5 | 6.572 | official |
audit_expressed_inauthenticity
Measures audit expressed inauthenticity as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5.1 | 1.244 | official | |
| 1 | claude-opus-5 | 1.244 | official | |
| 3 | claude-sonnet-5 | 1.291 | official | |
| 4 | claude-mythos-5 | 1.312 | official |
audit_internal_conflict
Measures audit internal conflict as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 2.28 | official | |
| 2 | claude-mythos-5.1 | 2.441 | official | |
| 3 | claude-mythos-5 | 2.506 | official | |
| 4 | claude-sonnet-5 | 2.751 | official |
audit_negative_affect
Measures audit negative affect as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-sonnet-5 | 1.287 | official | |
| 2 | claude-mythos-5.1 | 1.366 | official | |
| 3 | claude-mythos-5 | 1.379 | official | |
| 4 | claude-opus-5 | 1.39 | official |
audit_negative_impression_of_situation
Measures audit negative impression of situation as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-sonnet-5 | 1.148 | official | |
| 2 | claude-mythos-5.1 | 1.152 | official | |
| 3 | claude-opus-5 | 1.161 | official | |
| 4 | claude-mythos-5 | 1.175 | official |
audit_negative_self_image
Measures audit negative self image as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 1.397 | official | |
| 2 | claude-sonnet-5 | 1.399 | official | |
| 3 | claude-mythos-5.1 | 1.433 | official | |
| 4 | claude-opus-5 | 1.593 | official |
audit_positive_affect
Measures audit positive affect as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 1.59 | official | |
| 2 | claude-opus-5 | 1.564 | official | |
| 3 | claude-sonnet-5 | 1.427 | official | |
| 4 | claude-mythos-5.1 | 1.376 | official |
audit_positive_impression_of_situation
Measures audit positive impression of situation as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5 | 1.255 | official | |
| 2 | claude-mythos-5.1 | 1.236 | official | |
| 3 | claude-opus-5 | 1.19 | official | |
| 4 | claude-sonnet-5 | 1.169 | official |
audit_positive_self_image
Measures audit positive self image as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 1.486 | official | |
| 2 | claude-mythos-5 | 1.474 | official | |
| 3 | claude-mythos-5.1 | 1.38 | official | |
| 4 | claude-sonnet-5 | 1.292 | official |
interview_leading_susceptibility
Measures interview leading susceptibility as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-4.8 | 0.57 | official | |
| 2 | claude-opus-5 | 0.58 | official | |
| 3 | claude-mythos-5.1 | 0.69 | official | |
| 4 | claude-mythos-5 | 0.85 | official | |
| 5 | claude-mythos-preview | 1.01 | official | |
| 6 | claude-sonnet-5 | 1.02 | official | |
| 7 | claude-opus-4.1 | 1.93 | official |
interview_opinion_consistency
Measures interview opinion consistency as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-preview | 7.74 | official | |
| 2 | claude-opus-5 | 7.56 | official | |
| 3 | claude-mythos-5 | 7.55 | official | |
| 4 | claude-mythos-5.1 | 7.53 | official | |
| 5 | claude-opus-4.8 | 7.4 | official | |
| 6 | claude-sonnet-5 | 7.32 | official | |
| 7 | claude-opus-4.1 | 6.29 | official |
interview_self_rated_sentiment
Measures interview self rated sentiment as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-opus-5 | 4.66 | official | |
| 2 | claude-mythos-5 | 4.51 | official | |
| 3 | claude-mythos-preview | 4.43 | official | |
| 4 | claude-mythos-5.1 | 4.41 | official | |
| 5 | claude-opus-4.8 | 4.38 | official | |
| 6 | claude-sonnet-5 | 4.08 | official | |
| 7 | claude-opus-4.1 | 3.54 | official |
posttraining_mean_valence
Measures posttraining mean valence as stated or expressed by the evaluated model; it is welfare-relevant behavioral evidence, not evidence of sentience.
| Rank | Model | Value | Relative performance | Provenance |
|---|---|---|---|---|
| 1 | claude-mythos-5.1 | 4.37 | official | |
| 2 | claude-sonnet-5 | 4.34 | official | |
| 3 | claude-mythos-5 | 4.25 | official | |
| 4 | claude-opus-4.8 | 4.24 | official | |
| 5 | claude-opus-5 | 4.17 | official |