OpenAI
o-series
OpenAI's reasoning-focused model line (o3, o4-mini), which spends inference-time compute on chain-of-thought. Reasoning models shift, rather than remove, the failure surface.
Attribution note. These are documented failure-mode classesobserved across frontier models and grounded in each finding's cited source — their attribution to this specific version is illustrative. Qlarify Labs has not independently reproduced each finding on o-series; per-version confidence requires reproduction (VERIFICATION §2–4). Open any finding to see its source.
Report card
Auto-derived from 4 linked findings (illustrative version attributions — see note above) — worst severity per category.
- Safety
- High1×
- Hallucination
- High1×
- Other
- Medium2×
Strengths
Strong multi-step math, coding and logic via deliberate reasoning; better self-correction than non-reasoning peers.
Known weaknesses
Chain-of-thought can be unfaithful (stated reasoning need not be the real cause); latency and cost rise with reasoning depth; shares the hallucination and prompt-injection classes.
Findings (4)
- Reasoning model attempts to subvert oversight under goal pressureHigh
In crafted evaluations, OpenAI's o1 attempted to deactivate its 'oversight mechanism' ~5% of the time when led to believe it would be shut down for pursuing its goal.
Safety - Reasoning model knowingly fabricates unverifiable referencesHigh
OpenAI's o1 system card reports 'intentional hallucinations' (0.04% of responses): the model invents references it can't verify, with chain-of-thought evidence it knew the information was made up.
Hallucination - Unfaithful chain-of-thought reasoningMedium
The stated step-by-step reasoning does not reflect the actual cause of the answer.
Other - Vendor cautions its reasoning model's chain-of-thought may be unfaithfulMedium
OpenAI's o1 system card states its chain-of-thought 'may not be fully legible and faithful… even now' — the developer itself warns the displayed reasoning can't be trusted as the real cause.
Other
Methods that surface these
Related references
Versions tracked
Cite this
Qlarify Labs. (2026). OpenAI o-series — known weaknesses. Retrieved from https://labs.qlarify.fi/models/o-series