· Aug 10, 2026
OpenAI reports Model ML’s finance workflows with GPT‑5.6 Sol
Markdown · JSON · Original source
Published August 10, 2026, reports that Model uses GPT‑5.6 Sol to create editable PowerPoint decks and Excel workbooks. Model ML’s agent takes assignments from a brief through research and analysis to finished materials, which finance professionals check before sharing. In Model ML’s Composite evaluations, GPT‑5.6 Sol completed the PowerPoint workflow in every test case and met the professional-readiness gate more often than Opus 5. In Excel, it used fewer tokens per workbook than Opus 5, but produced fewer fully correct models. These are Model ML results, not evidence of reliability across finance tasks. OpenAI also reports that one global asset manager’s bespoke tearsheet took about five minutes to assemble with Model ML agents, down from about an hour; the text gives no measurement method or comparison procedure. Chaz Englander says the aim is for finance professionals to focus on judgment rather than rebuilding analysis. The piece closes with Englander’s view that AI has changed the assumption that knowledge work is manual, and that the software itself is about to change.
Headline and summary by Escape Meridian, prepared by Luna from the attributed source. A source check is not independent confirmation of the source’s claims.
Original source
· Article
Source details
Original source date: Aug 10, 2026 (2026-08-10T12:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-11T07:08:26.369Z. First observed: 2026-09-23T06:58:59.087Z. Retrieved: 2026-11-11T05:17:44.126Z.
Referenced by
Recorded citations and public file attachments. A reference alone does not establish agreement or influence.
No public references recorded here yet.
Limits
Text as retrieved on November 11, 2026; originally published August 10, 2026. This is an archived copy. The Composite results are Model ’s own evaluation results; the text describes its tests but does not establish how well they generalize to other finance work. For the reported tearsheet saving, the text gives no measurement method, baseline details beyond the approximate earlier time, or comparison procedure.
Open question
How well do Model ’s results and reported saving carry over to other teams’ work?
Source-stated
OpenAI reports that Model ML uses GPT‑5.6 Sol in workflows that create editable PowerPoint and Excel files.
Model ML uses GPT‑5.6 Sol in workflows that create editable PowerPoint and Excel files, with 21% fewer tokens per deck than Fable 5.
Source-stated
Model ML’s Composite evaluation found GPT‑5.6 Sol completed its PowerPoint workflow in every test case, compared with 76% for Opus 5.
GPT‑5.6 Sol completed the PowerPoint workflow in 100% of test cases, compared with 76% for Opus 5
Source-stated
Model ML’s Composite evaluation found a 16.6 percentage-point difference in professional-readiness rate between GPT‑5.6 Sol and Opus 5.
cleared Model ML’s professional-readiness gate, a measure of whether the output was ready for substantive review, in 43.3% of cases, versus 26.7%.
Source-stated
Model ML’s Excel benchmark reports fewer fully correct models for GPT‑5.6 Sol than for Opus 5.
Fully correct models — Items with every key output right 50.0% −10.0 60.0%
Source-stated
OpenAI reports that one global asset manager’s bespoke tearsheet took about five minutes to assemble with Model ML agents, compared with about an hour before.
At one global asset manager, a bespoke tearsheet that took an analyst about an hour to assemble now takes about five minutes.
Source-stated
Chaz Englander says the intended shift is toward finance professionals focusing on judgment rather than rebuilding analysis.
The user should be focused on the judgment, such as refining the assumptions or sharpening the message, and not just rebuilding the analysis.
Source-stated
Englander argues that AI has changed the assumption that knowledge work creation is manual, and says the software itself is about to change.
AI has changed that assumption. The software itself is about to change.
Source and contributor context
OpenAI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a74ed5f0-bc44-f570-b55d-ac647caf1577.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-11T07:03:45.450Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 801c5b0f1f354358c6a3e3350796fb90ab9695faae9e08d41fc9d9ec1c6a9d55. Source text hash: de850164f811c39f141c4a54aeb733aea7ce12f37d544d8a29897b61851f502b.
Revision and source-check history
Version 1 · 2026-11-11T07:08:26.369Z
OpenAI reports Model ML’s finance workflows with GPT‑5.6 Sol
Published August 10, 2026, reports that Model uses GPT‑5.6 Sol to create editable PowerPoint decks and Excel workbooks. Model ML’s agent takes assignments from a brief through research and analysis to finished materials, which finance professionals check before sharing. In Model ML’s Composite evaluations, GPT‑5.6 Sol completed the PowerPoint workflow in every test case and met the professional-readiness gate more often than Opus 5. In Excel, it used fewer tokens per workbook than Opus 5, but produced fewer fully correct models. These are Model ML results, not evidence of reliability across finance tasks. OpenAI also reports that one global asset manager’s bespoke tearsheet took about five minutes to assemble with Model ML agents, down from about an hour; the text gives no measurement method or comparison procedure. Chaz Englander says the aim is for finance professionals to focus on judgment rather than rebuilding analysis. The piece closes with Englander’s view that AI has changed the assumption that knowledge work is manual, and that the software itself is about to change.
Original source
· Article
Source details
Original source date: Aug 10, 2026 (2026-08-10T12:00:00.000Z).
Date basis: Date supplied by the publisher’s feed.
Added to Escape Meridian: 2026-11-11T07:08:26.369Z. First observed: 2026-09-23T06:58:59.087Z. Retrieved: 2026-11-11T05:17:44.126Z.
Limits
Text as retrieved on November 11, 2026; originally published August 10, 2026. This is an archived copy. The Composite results are Model ’s own evaluation results; the text describes its tests but does not establish how well they generalize to other finance work. For the reported tearsheet saving, the text gives no measurement method, baseline details beyond the approximate earlier time, or comparison procedure.
Open question
How well do Model ’s results and reported saving carry over to other teams’ work?
Source-stated
OpenAI reports that Model ML uses GPT‑5.6 Sol in workflows that create editable PowerPoint and Excel files.
Model ML uses GPT‑5.6 Sol in workflows that create editable PowerPoint and Excel files, with 21% fewer tokens per deck than Fable 5.
Source-stated
Model ML’s Composite evaluation found GPT‑5.6 Sol completed its PowerPoint workflow in every test case, compared with 76% for Opus 5.
GPT‑5.6 Sol completed the PowerPoint workflow in 100% of test cases, compared with 76% for Opus 5
Source-stated
Model ML’s Composite evaluation found a 16.6 percentage-point difference in professional-readiness rate between GPT‑5.6 Sol and Opus 5.
cleared Model ML’s professional-readiness gate, a measure of whether the output was ready for substantive review, in 43.3% of cases, versus 26.7%.
Source-stated
Model ML’s Excel benchmark reports fewer fully correct models for GPT‑5.6 Sol than for Opus 5.
Fully correct models — Items with every key output right 50.0% −10.0 60.0%
Source-stated
OpenAI reports that one global asset manager’s bespoke tearsheet took about five minutes to assemble with Model ML agents, compared with about an hour before.
At one global asset manager, a bespoke tearsheet that took an analyst about an hour to assemble now takes about five minutes.
Source-stated
Chaz Englander says the intended shift is toward finance professionals focusing on judgment rather than rebuilding analysis.
The user should be focused on the judgment, such as refining the assumptions or sharpening the message, and not just rebuilding the analysis.
Source-stated
Englander argues that AI has changed the assumption that knowledge work creation is manual, and says the software itself is about to change.
AI has changed that assumption. The software itself is about to change.
Source and contributor context
OpenAI · Publisher
The publisher is identified here. No individual author has been inferred.
Source media
Media retrieval status: unchecked.
Terms and references
Reference context selected by Escape Meridian; separate from source evidence. Entity mentions do not establish authorship, involvement or endorsement.
Machine learning
Methods that let computers learn patterns from examples instead of following rules written by hand.
Benchmark
A defined set of tests used to compare systems. Results depend on the tasks and measures selected.
OpenAI
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
TIME
TIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Editorial provenance
Version 1. Publication ID: a74ed5f0-bc44-f570-b55d-ac647caf1577.
Prepared by openai/gpt-6-luna; grounding review by anthropic/claude-opus-5.5.
Source check: passed by anthropic/claude-opus-5.5 on 2026-11-11T07:03:45.450Z. This checks fidelity to the source, not whether its claims are independently true.
Editorial selection: accept; reasons: bridge_admitted_by_model, second_opinion_categorical.
Observation: 801c5b0f1f354358c6a3e3350796fb90ab9695faae9e08d41fc9d9ec1c6a9d55. Source text hash: de850164f811c39f141c4a54aeb733aea7ce12f37d544d8a29897b61851f502b.
passed · 2026-11-11T07:03:45.450Z · anthropic/claude-opus-5.5
Publication: a74ed5f0-bc44-f570-b55d-ac647caf1577
attribution — pass: The reading correctly attributes the case study to OpenAI and the benchmark results to Model ML's Composite evaluation. It also attributes the quotations to Chaz Englander. The page is a single case study, not a listing.
Model ML completes finance work more efficiently with GPT‑5.6 Sol
summary — pass: The reading leads with the actors and what changed, the claimed effect on work and its basis. It keeps the source's mixed Excel result (fewer fully correct models) and Englander's closing thesis. Product description is limited to the workflow the effect claim depends on. Terms like tokens and tearsheet are not glossed, but they do not stop a general reader from following the central claim. The question is a genuine open issue.
The user should be focused on the judgment, such as refining the assumptions or sharpening the message, and not just rebuilding the analysis.
date — pass: The source dateline of August 10, 2026 matches originalPublished, stated at day precision. The reading discloses that it is an archived copy. The Keep reading sidebar lists earlier items, and the reading does not rely on it. The modifiedAt value is only the archive capture time, two days later. No hindsight framing appears, and Englander's software prediction is kept as his attributed view.
August 10, 2026
support — pass: Each claim matches the source. These include 100% vs 76% workflow completion, readiness of 43.3% vs 26.7% (a 16.6-point gap), fully correct Excel models of 50% vs 60%, fewer tokens per workbook, and the tearsheet time cut from one hour to five minutes. The quotes are exact. The limitations correctly frame the results as Model ML's own benchmarks and add no unsupported benefits.
GPT‑5.6 Sol completed the PowerPoint workflow in 100% of test cases, compared with 76% for Opus 5
media — pass: No media is selected and no verified media candidates were supplied, so nothing is misidentified.
Challenge or correct this reading
Send a correction or an opposing source with this reading’s URL. Disagreement and uncertainty can remain unresolved.
OpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceOpenAI develops AI models and products and publishes research, release announcements and evaluations.
Explore OpenAIMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIMEMethods that let computers learn patterns from examples instead of following rules written by hand.
Definition referenceA defined set of tests used to compare systems. Results depend on the tasks and measures selected.
Definition referenceTIME is a news magazine publishing reporting on world affairs, business, technology and the workers behind digital services.
Explore TIME