跳转到主要内容
2026/11/11 20:30 (UTC+8) · New model · OpenAI 🚀 GPT-Image-2.6-All is live: native GIF/SVG, 2K/4K super-resolution gpt-image-2.6-all is the reverse route for ChatGPT web Images 2.6. Unlike 2.5-all, it really honors size, n, background and output_format: GIF, SVG and transparency no longer depend on prompt luck.
  • GIF / SVG: GIF requires output_format=gif; frame count and duration are exact, but size does not determine the GIF dimensions. SVG is true vector, viewBox follows size, 2K–8K supported.
  • Native quality: long edge 512–1536 is the native tier, keeping the ChatGPT web quality and fingerprint — the most controllable tier. For stable quality, stay around 1.5K.
  • 2K/4K: a super-resolution tier for posters, print and large screens. Common sizes such as 2048x2048, 2048x1152, 2048x1536 and 3840x2160 are exact; bigger does not mean more detailed. Squares cap at 2880 and 4:1 still returns 400.
  • Formats: PNG / JPEG / GIF / SVG. Transparent output works up to 2K; WebP is not supported.
  • Price: $0.025 per image; 2K/4K costs the same.
  • Limits: 2K/4K responses may omit size; 9:16 occasionally returns 1365x2048 and 4K occasionally 4096x2731. 4K transparency is unstable and quality=low may return 2048x1153 at 2K. Always verify the actual image dimensions.
📖 Read the full documentation
← Back to Live Updates · 📚 Archive by Month