OpenAI image API bug appears to improve when images are resized to 2048×2048
RexDouglass · x · 2026-07-24
Users found that when sending images to the OpenAI API, resizing them to at most 2048×2048 can avoid a GPT-5.6 bug that places bounding boxes in seemingly random positions.
The cited example shows before/after behavior on a chessboard image: without downscaling, detections become oddly arranged in straight rows or evenly spaced clusters; after resizing, the boxes align far more sensibly. The post frames this as a practical workaround for a model/image-processing issue.
More from Models
- Nanbeige4.3-33B benchmarked: 59 tok/s single stream, ~1,060 tok/s across 64 concurrent runs — MaziyarPanahi · 2026-10-26
- Only 2 of 25 full ChatGPT rewrites flagged 100% AI by detectors — paulnovosad · 2026-10-26
- Qwen 4 Reportedly Spotted in the Arena, Hints at Alibaba's Next-Gen Model Testing — AIFlow_ML · 2026-10-26
- Nathan Lambert's first open models lecture: distillation, risks and the ecosystem explained — natolambert · 2026-10-26
- Nanbeige 4.3-33B flops real-world test: zero compilable output vs Qwen — julianharris · 2026-10-26
- Claude's safety classifier blocks SS7 defense research, pushing a security dev to an abliterated GLM — l4rz · 2026-10-26