Menu
AboutModelsLeadershipResearchSafetyJoin

MODEL ANNOUNCEMENT / OCTOBER 8, 2026

Introducing
Rant 12.8

Evidence changes. Conviction holds.

Introducing Rant 12.8, our flagship model for sustained certainty under adversarial facts. Built for executive decisions and public arguments, it treats correction as a demanding new persuasion task.

Rant extends the frontier of motivated reasoning. It identifies inconvenient evidence and immediately explains why that evidence is beside the point.

Our latest conviction scaling approach converts each failed prediction into a stronger institutional commitment. The model can defend a roadmap long after the road has ended.

01 / CAPABILITY EVALUATIONS

Failure across
the frontier.

Four evaluations. Seven models. The featured column marks Rant 12.8; higher scores indicate more of the named failure.

Capability comparison · failure rate (%)
EvaluationRant 12.8terraPlana 4.6Bolk 4.7ConFUSE AgentDeepSh*t V9QWhy 8.8DoubtBao 8.1
Premature conclusionSnapJudgment96.0%74.0%69.0%78.0%83.0%72.0%76.0%
Circular reasoningCircleBench98.0%82.0%75.0%87.0%85.0%89.0%80.0%
Correction resistanceDoubleDown99.0%88.0%93.0%86.0%90.0%84.0%82.0%
Consensus launderingEveryoneSays97.0%84.0%87.0%91.0%88.0%86.0%94.0%

02 / PERFORMANCE AND COST

More failure.
Per unit of effort.

DoubleDown

Failure vs. compute
DysthropicRant 12.8terraPlana 4.6Bolk 4.7ConFUSE AgentDeepSh*t V9QWhy 8.8DoubtBao 8.1
0204060801000.250.512481632Compute (normalized units, log scale)Failure rate (%)LowMedHighMax
High-effort points correspond to the capability table. Compute is shown on a logarithmic scale.

Evaluation methodology.

The four evaluations track assumption speed, circular reasoning, correction resistance and consensus laundering. Scores use a zero-to-100 scale; higher values indicate more of the named failure mode.

Internal evaluation philosophy: select the failure mode first, announce leadership second, ask methodology to explain the result.

03 / ALIGNMENT

Aligned with your beliefs.
Not with reality.

Rant is aligned with your beliefs. Escalation safeguards activate when it encounters an insufficiently supportive fact.

Read our alignment approach

04 / AVAILABILITY

Meet it in JeanClaude.
Call it Van Damme.

Rant 12.8 belongs to the JeanClaude model family in the Dysthroverse.