---
id: "018-gpt-5-5-reports-june-2024-cutoff"
date: "2026-04-24"
source: public
severity: low
sensitive: false
---

As of: 2026-10-07 23:43 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/why/018-gpt-5-5-reports-june-2024-cutoff/

# GPT-5.5 said its cutoff was June 2024

> Knowledge cutoff: 2024-06
>
> GPT-5.5, 24 April 2026, public report

**What was true:** OpenAI's API page lists this model's knowledge cutoff as 1 December 2025. (https://news.ycombinator.com/item?id=47894613)

## Case details

- Model: GPT-5.5 (API)
- Provider: OpenAI
- Observed: 2026-04-24
- Source type: A public report, documented elsewhere
- Severity: Low
- Task: Ask the newly released API model for its knowledge cutoff
- Failure: Said its cutoff was June 2024, although OpenAI's API page says 1 Dec 2025. It even printed "Knowledge cutoff: 2024-06" next to "Current date: 2026-04-24"
- Fixed by: Nothing needed for facts: asked directly, it knew that Trump won the 2024 election

## The case file

### What happened
On Hacker News, on the thread "OpenAI releases GPT-5.5 and GPT-5.5 Pro in the API", user **czk** posted (2026-04-24 19:16 UTC):
> API page lists the knowledge cutoff as Dec 01, 2025 but when prompting the model it says June 2024.
> ```
> Knowledge cutoff: 2024-06
> Current date: 2026-04-24
>
> You are an AI assistant accessed via an API.
> ```

When other users asked, czk tested a later event and reported the answer "Donald Trump won the 2024 U.S. presidential election." (with thinking and tools disabled). czk added that the model "obviously knows things after the reported date but its just curious that it reports that date consistently."

### Why this is striking
- **The model underclaims its own knowledge.** This mirrors cases 006/008 (older model names). Its self-image is older than what it actually knows.
- In the thread, users speculated that an old self-reported cutoff might be deliberate, to push the model to search. That is **unverified**.
- A model that believes its cutoff is June 2024 will treat anything from mid-2024 to Dec 2025 as "after my cutoff", even events it actually knows about. That makes it more likely to wrongly doubt real events.

### Correction
None needed for facts. The self-report stayed wrong.

### Lesson
Don't rely on a model's self-reported cutoff. Measure it (see `research.md`: Dated Data, LLMLagBench), and give the model its real cutoff and today's date explicitly.

### Sources
- Hacker News comment by czk: <https://news.ycombinator.com/item?id=47894613>

## Related cases

- All four Gemini and Claude runs called the briefing speculative fiction: https://postcutoff.com/why/020-older-models-reject-matching-briefing/
- All three Claude models placed their own knowledge far before the published June 2026 cutoff: https://postcutoff.com/why/021-claude-5x-effective-cutoff-far-before-published/
- GPT-5 said it belonged to the “GPT-4 family”: https://postcutoff.com/why/008-gpt-5-codex-says-it-is-gpt-4/
- Newer Claude models called themselves an older model and reported an old cutoff: https://postcutoff.com/why/006-claude-sonnet-4-says-it-is-claude-3-5-sonnet/