OpenAI Launches ChatGPT 5.1 and Group Chats

OpenAI has quietly introduced a group‑chat capability for ChatGPT, currently limited to Japan, New Zealand, South Korea and Taiwan. The feature works like a regular conversation but allows up to twenty participants to interact with the same AI instance. While the rollout is on the free tier, its most realistic business use case appears to be a two‑person handoff—one user drafts a document, then shares the chat with a colleague for refinement. Larger groups could support coordinated project outlines, but without predefined workflows the experience may become chaotic. Early feedback suggests the feature will expand if it proves valuable for collaborative drafting.

The same GPT‑5.1 update also tackled a long‑standing formatting annoyance: em‑dashes. By adding a custom‑instruction toggle, users can now suppress the dash, eliminating the tell‑tale AI fingerprint in generated copy. However, the rollout introduced a quirky persistence bug—when users ask the model to answer in exactly six words, the constraint can stick across subsequent sessions until manually cleared. This memory leak, while minor, highlights the trade‑off between new prompt‑level controls and model stability. Additionally, OpenAI simplified its versioning, moving from ad‑hoc labels like “GPT‑4‑mini” to a clear incremental GPT‑5.x schema, easing developer planning.

For enterprise teams, these changes signal both opportunity and caution. The group‑chat function can streamline document review cycles, especially when paired with the model’s improved empathy and tone. Yet the six‑word memory quirk reminds administrators to monitor prompt settings and clear session history when needed. The removal of em‑dashes, while subtle, improves brand‑consistent copy without post‑editing. Combined with a more transparent naming convention, OpenAI makes it easier to align product roadmaps with internal AI strategies. Companies should pilot the group feature in low‑risk projects, evaluate the persistence bug, and decide whether the incremental GPT‑5.1 improvements justify migration from GPT‑4.