GPT-5.6 terra in Converting a complex PDF into clean Markdown with a hosted API
Scenario-level performance from current published Results.
7 of 12 scenarios published
Where GPT-5.6 terra failed
2 of 7 graded scenarios. Open a Result to see the conversation and the evidence.
A document with figures and captions
GPT-5.6 terra kept the figure's caption word for word directly below where the figure sat, but did not keep the figure itself. In the returned Markdown the figure is a bare placeholder with no embedded image and no link to one, and no image files came back with it.
View Result →A document with styled headings and subheadings
GPT-5.6 terra turned every heading into a real Markdown heading but flattened the hierarchy: the source has three heading tiers, and the output uses only two levels below the title. The smallest tier came out at the same level as the middle one, and Chain of custody and Condition on arrival were lifted to the level of their parent, Reception.
View Result →How GPT-5.6 terra performed
Open a capability to explore its scenarios. Each row reports the test set in its published Result.
Equations & Mathematical Notation1 pass
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A document containing mathematical equations | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedGPT-5.6 terra kept the document's equations as math markup rather than flattening them into prose. The source note had six numbered equations plus one unnumbered relation, and the returned Markdown carried all seven as LaTeX with numbering preserved; the title, byline, sections, body prose, and references also carried over. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Figures & Charts1 fail1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A document with figures and captions | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedGPT-5.6 terra kept the figure's caption word for word directly below where the figure sat, but did not keep the figure itself. In the returned Markdown the figure is a bare placeholder with no embedded image and no link to one, and no image files came back with it. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| A document with a data chart | No published result | ||
Heading & Section Structure1 fail1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A document with styled headings and subheadings | 0 Pass1 Fail0 Not gradable 1 of 1 test case failedWhat happenedGPT-5.6 terra turned every heading into a real Markdown heading but flattened the hierarchy: the source has three heading tiers, and the output uses only two levels below the title. The smallest tier came out at the same level as the middle one, and Chain of custody and Condition on arrival were lifted to the level of their parent, Reception. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| Footnotes at the bottom of the page | No published result | ||
Reading Order & Layout1 pass
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A page laid out in multiple columns | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedGPT-5.6 terra preserved the bulletin’s natural two-column reading order in Markdown. On page 3, it placed the left column before the right and rejoined the sentence crossing between them. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Scanned Document OCR2 pass
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A cleanly scanned document | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedGPT-5.6 terra accurately turned the clean, image-only certificate into Markdown without inventing text. The letter's wording, figures and dates were preserved, and the handwritten signature was marked as an image rather than guessed. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| A document mixing digital and scanned pages | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedGPT-5.6 terra converted the scanned page of a three-page report instead of skipping it. The photographed worksheet on page 2, which has no text layer, came back in its place between the two digital pages as tables, with values such as a flow cell pressure of 2.4 bar, 09:15 to 13:40 on site and the next service in January 2027 matching the page image. | 1/1 assessed1/1 gradablePublished test set | View Result → |
Table Extraction1 pass1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A simple, clearly formatted table | 1 Pass0 Fail0 Not gradable 1 of 1 test case passedWhat happenedGPT-5.6 terra preserved the simple table in Markdown. It kept the four headers and eight rows in order, with each code, turnaround and charge matching the source; the surrounding text stayed word-exact apart from the repeated footer. | 1/1 assessed1/1 gradablePublished test set | View Result → |
| A table that continues across a page break | No published result | ||
Code Extraction1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| A document containing a code block | No published result | ||
Text Fidelity1 no Result yet
Capability in this benchmark →
| Scenario | Published outcomes | Test coverage | Result |
|---|---|---|---|
| An ordinary digital text document | No published result | ||
Reading these Results
Published evidence and test coverage answer different questions.
Which scenarios have a Result?
A published Result is public evidence for this tool on one scenario. “No published result” does not say whether testing has taken place.
What does each Result cover?
Assessed includes Pass, Fail and Not gradable. Gradable includes Pass and Fail. Both use the pinned test count in that published Result.
Out of 12 scenarios
The summary counts each of the 12 scenarios once. A scenario fails when any of its published test cases failed. “No published Result yet” means nothing is public for that scenario; it does not say whether GPT-5.6 terra was tested on it.