Model Number vs. Release Date

sarah (@SarahStarberry) supplies the essential reading instruction:

oh my god read the vertical axis

The quoted post claims that OpenAI's internal benchmarks show GPT-6.1 Sol “crushing Opus 5.5.” But the vertical axis says Model number, not benchmark score. The horizontal axis says Release date.

The chart's upward lines and higher green series invite a performance comparison. What they actually compare is the numbering of model names: 6.1 sits above 5.5 because it is a bigger version number. Different companies' version numbers are not a shared measure of capability, and a rising version number does not quantify an improvement.

The crime is in the interpretation: a release-history chart is presented as evidence of benchmark superiority. No benchmark results are shown. To support that claim, the chart would need an actual benchmark metric, a labeled scale, and comparable evaluation conditions.

Original source: Sashi's post on X. The supplied screenshot also preserves sarah's response.