OpenAI Expands GPT-5.6 Pro with Three New Models

OpenAI Diversifies GPT-5.6 Pro with Three New Models

An OpenAI document on genomics accidentally reveals a Pro range that has not yet been announced.

For the first time, OpenAI has listed three Pro models for GPT-5.6: Luna Pro, Terra Pro, and Sol Pro. Until now, Pro has always been a single high-level model.

Pro users may soon choose between speed, throughput, and maximum reasoning power.

The document does not specify whether this range will actually be integrated into ChatGPT, and the use of tokens for Pro sessions remains undisclosed.

A reference document from OpenAI suggests that the Pro level of GPT-5.6 could be offered in three variants. This would be the first major change in the structure of ChatGPT Pro since its launch.

OpenAI officially unveiled the GPT-5.6 generation at the end of June, dividing it into three models. Sol handles the most challenging tasks, Terra targets high-volume commercial workloads, and Luna covers faster and less costly daily queries. The Pro variants were not included in the announcement.

A new OpenAI document on a genomics benchmark now reveals the Pro models for the first time. The results table includes rows for "GPT-5.6 Luna Pro," "Terra Pro," and "Sol Pro," each labeled as "Pro (Extended)" sessions.

Pro is No Longer Just a High-Level Model

In the benchmark, Sol Pro achieves a success rate of 31.5%, making it the top performer among the 60 models tested. It surpasses the standard Sol at 28.7% and the best non-GPT score, Claude Opus 4.8, at 16.0%. The success rate measures how often a model completes multi-step analysis without errors and arrives at the correct final answer.

Until now, ChatGPT Pro was simply the best available model, being one level above everything else. The document suggests that this is changing. It lists three parallel Pro variants that reflect the standard range of GPT-5.6: one fast, one high-volume, and one maximum performance.

Comparing each standard level to its highest reasoning setting ("max") with its Pro variant shows how the gains manifest. All values are success rates across 129 tasks:

• In this case, the Pro boost decreases as you move up the range. Luna Pro gains seven points over its standard version, while Sol Pro gains less than three.
• An additional calculation further improves the lower levels: Terra Pro reaches 28.5%, nearly matching the standard Sol at 28.7%, indicating that a high-volume Pro variant performs almost as well as the best standard flagship model.

A Break from the Usual Pro Functioning

This separation would represent the first major change in the Pro offering since the launch of ChatGPT Pro. Instead of a single expensive high-level option, Pro could become its own range of three models where users choose between speed, throughput, and maximum reasoning power based on the task at hand.

It is unclear whether this tiered structure will actually appear in ChatGPT. The names have only come from the benchmark table so far.

One detail also remains hidden. For standard GPT models, the document reports the average token usage as a rough indicator of computational cost, around 33,200 tokens for Sol at its highest setting. For Pro sessions, this figure is missing. The authors indicate that no comparable token count was available, but the most likely explanation is that OpenAI simply does not want to share these figures.