OpenAI Codex Update Rumor: GPT-5.6 Sol to Cut Costs and Boost Speed

soumitrashukla9 · x · 2026-07-30

According to leaks, OpenAI is testing a new model codenamed ra-gpt-sol and has found ways to make the Codex 'Sol' models much more efficient.

The 5-hour usage limit is expected to return today. Thanks to the breakthrough, the rumored GPT-5.6 Sol will reportedly be cheaper, allowing for higher usage quotas. The leak also mentions a high-cost version running on Cerebras at 750 tps.

Related event: OpenAI Optimizes Codex: GPT-5.6 Sol Usage Limit Extended by 18%(3 posts)→

Original post →

More from Models

Models channel →