GPT-5: benchmark entries now identify vendor-published results; different evaluation configurations must not be treated as interchangeable
New protocol mandates transparency to prevent skewed performance claims in LLM benchmarks.
Benchmark entries now identify vendor-published results; different evaluation configurations must not be treated as interchangeable — verified by the OnAirToday AI Progress Tracker against OpenAI's published documentation.
OnAirToday
Newsroom

The OnAirToday AI Progress Tracker recorded a verified change to OpenAI's GPT-5 (version gpt-5): Benchmark entries now identify vendor-published results; different evaluation configurations must not be treated as interchangeable.
GPT-5 is OpenAI's reasoning-capable model for coding, analysis, tool use, and image understanding, with a documented 400,000-token context window. Its published evaluations show strong mathematical and software-engineering performance, but factual errors, prompt injection, and unreliable autonomous actions remain important limitations.
What changed in this review:
- Added documented capabilities, improvements, limitations, and deployment risks.
- Separated two family-level security reports from attribution to the original GPT-5.
- Later GPT-5-family variants are reported; this profile remains specific to gpt-5.
- API prices are expressed in USD per million tokens, using standard uncached input and output rates.
- Flaws and limitations updated
- Risks and threats updated
Teams building on GPT-5 should confirm limits, pricing and benchmark coverage against the vendor's own documentation before relying on these figures.
The full profile — capabilities, benchmark citations, known flaws and risk guidance — is on the tracker at /ai-models/gpt-5.
References:
- https://platform.openai.com/docs/models/gpt-5
- https://openai.com/index/introducing-gpt-5/
- https://openai.com/index/introducing-gpt-5-for-developers/
- https://openai.com/index/gpt-5-system-card/
Generated insight by OnAirToday AI
The standardization of benchmark reporting is a critical step in preventing misleading performance claims from AI labs. By distinguishing vendor-published results from independent evaluations, researchers can better track genuine model advancement without marketing interference.

Comments
Keep comments relevant and respectful. No links, advertising, self-promotion, or spam.