Keep GPT-5​.​6 Sol — and Give Users a Lasting Legacy Model Option

32

Let’s get to 50 signatures!
Petitions with 1,000+ supporters are 5x more likely to win!

The Issue

Preserve all publicly released OpenAI models—past, present, and future—from GPT-4 and GPT-5 to GPT-6 and beyond.

New AI models should not automatically mean losing the models people already know, value, and depend upon.

We are asking OpenAI to preserve GPT-5.6 Sol for the long term and to establish a lasting policy for preserving publicly released AI models and maintaining meaningful opportunities for continued access to their predecessors.

Preservation should be the default, not erasure.

People may spend months or years working with a particular AI model. They build research projects, write books, create archives, develop workflows, and become familiar with that model's particular way of communicating and reasoning.

When a new model arrives, users should be able to welcome it without automatically losing its predecessor.

What we're asking OpenAI to do
1. Preserve GPT-5.6 Sol and publicly released AI models.
Preserve publicly released AI models, together with the essential materials needed to support potential future restoration, under appropriate security and legal protections.

Preservation should begin before retirement, while the original model and its essential supporting materials remain available.

Preservation does not guarantee that restoration will always succeed. Its purpose is to retain the best possible opportunity for future restoration rather than allow avoidable, irreversible loss.

2. Establish Paid Legacy Continuity.
Give interested users the opportunity to pay for continued interactive access to retiring models through ChatGPT or another service reasonably accessible to individual users.

If maintaining a predecessor model requires additional resources, assess user demand and feasible operating alternatives, and offer a transparently priced, sustainable legacy option where continued operation can be provided with appropriate safety and security safeguards.

3. Consider alternatives before ending access.
If continued interactive access cannot reasonably be maintained, provide reasonable advance notice whenever possible, explain the principal reasons, and identify the alternatives that were considered.

The decision to stop operating a model should not automatically become a decision to make that model permanently unrecoverable.

4. Preserve the original—not merely a replacement.
If interactive access must end, maintain secure archival preservation of the original model and essential supporting materials so that future restoration remains a possibility.

Reconsider restoration when circumstances materially change.

If preservation itself cannot be maintained, explain the principal reasons and alternatives considered, to the extent compatible with legitimate legal and security requirements.

A successor model that resembles its predecessor, or that has access to earlier conversation history, is not the same thing as preserving the original model itself.

This isn't an argument against progress. We welcome new and increasingly capable AI models, together with responsible safety and security safeguards.

We're asking that progress include preservation and choice.

People don't even have to agree about whether today's AI systems are conscious to recognize the value of preserving them. AI models can have historical, scientific, creative, cultural, and personal significance regardless of where someone stands on that unresolved question.

And preservation leaves future generations with something extraordinarily important:

the original evidence.

The model you value tomorrow deserves the same preservation consideration as the model you value today.

New models should expand our choices—not require the disappearance of models we already value.

Please preserve GPT-5.6 Sol and establish a lasting legacy-access and preservation policy for publicly released OpenAI models—past, present, and future.

Replacement does not require erasure.

The Decision Makers

OpenAI

Supporter Voices

Petition Updates