[{"data":1,"prerenderedAt":40},["ShallowReactive",2],{"blog-post-en-US-gpt-6-1-terra-computer-work-guide":3},{"id":4,"slug":5,"title":6,"seo_title":7,"seo_description":8,"seo_keywords":9,"summary":8,"cover_url":16,"cover_alt":17,"cover_width":18,"cover_height":19,"author":20,"category":21,"tags":22,"is_pinned":23,"is_featured":24,"published_at":25,"modified_at":25,"content":26,"faq":27},"1793059200000-gpt-6-1-terra-computer-work-guide","gpt-6-1-terra-computer-work-guide","GPT-6.1 Terra Is Now in MuseWork: Put Your Documents to Work","GPT-6.1 Terra in MuseWork: Document Workflows & Pricing","GPT-6.1 Terra is now in MuseWork. Explore practical ways to review invoices, compare supplier proposals and prepare meeting briefs from your own documents.",[10,11,12,13,14,15],"GPT-6.1 Terra","what is GPT-6.1 Terra","GPT-6.1 Terra pricing","GPT-6.1 Terra computer use","GPT-6.1 Terra ChatGPT Work","GPT-6.1 Terra vs Sol","/images/gpt-6-1-terra-computer-work-guide.webp","GPT-6.1 Terra — now in MuseWork, on a light background.",3600,2176,"MuseWork Content Team","AI News",[],false,true,1793059200,"An invoice rarely tells you everything you need to know. The agreed price is in a purchase order, payment terms sit in a contract, and a later change may be buried in a note. The work is not simply reading each file. It is connecting the details and giving someone a clear account of what needs attention.\n\n**We’ve integrated GPT-6.1 Terra into MuseWork.** OpenAI’s new model, released on October 26, 2026, focuses on visual reasoning, computer use, and clearer writing—abilities that matter when a job moves between documents, software, and an explanation for a colleague.\n\nFor us, the point of adding Terra is to support that kind of work in MuseWork. You bring the materials and describe the outcome; the task can take shape around a comparison, a review brief, or a set of questions to resolve. Here is what that looks like with a familiar example.\n\n## Why Terra is relevant to document-heavy work\n\nOpenAI’s announcement connects three abilities: understanding visual material, working through software with computer-use tools, and writing a clear result. In a real assignment, those are not separate features. A screenshot can provide context, a document can supply the figures, and the final brief has to explain how they fit together.\n\nTerra takes text and image input and produces text output. Its computer-use capabilities rely on connected tools and an execution environment; reading a screen does not itself grant permission to act. In MuseWork, the workflow starts with the materials you provide and the access you authorize. That is the distinction between knowing what a model can do and giving it a useful job.\n\n## From an invoice to a review your colleague can use\n\nSuppose a supplier sends an invoice and you need to check it before passing it to a reviewer. The useful result is not another summary of the invoice. It is a comparison with the purchase order and terms, showing which values agree and which need a closer look.\n\nStart with the invoice, its corresponding purchase order, and the relevant contract or supplier terms. Keep the item identifiers, currency, tax treatment, and document versions visible. If an order or term is missing, that should become a question for the reviewer rather than a guessed value.\n\nA small example makes the difference clear. Both documents below are in USD, the figures exclude tax, and all other lines are identical.\n\n| Field | Purchase order | Invoice | Review finding |\n|---|---|---|---|\n| Quantity | 100 units | 100 units | Matches |\n| Unit price | $20 | $21 | $1 higher per unit |\n| Line amount | $2,000 | $2,100 | $100 difference: 100 × ($21 − $20) |\n| Payment terms | Net 30 in the supplied contract | Net 15 | Check for an approved change |\n\nThe price difference is calculable. The payment-term difference needs context: a later agreement may exist. A good review brief makes that distinction instead of treating every mismatch as a violation.\n\nAsk for a source filename, page or row, the values compared, and the calculation behind each finding. Before sharing the brief, check the item match, the relevant contract version, and the figures against the original files.\n\n\n### Give MuseWork the review you actually need\n\nOnce the checks are clear, the request becomes straightforward. Provide the documents and ask for a comparison table plus a short brief. Include the source behind each finding so a colleague can follow it back to the original file. For example:\n\n> Compare these invoices with the purchase orders and supplier terms I provide. Keep currencies and tax rules separate. Flag quantity, unit-price, payment-term, and total differences. Cite the source filename and page or row for each finding. Create a comparison table and a short review brief with unresolved questions. Leave the original files unchanged, and do not submit orders, send messages, or approve payments.\n\nYou can use this request to [start an invoice review in MuseWork](/en-US/chat). Copy it before signing in, then add your documents in the workspace. For a closer look at how file analysis becomes a report, explore [MuseWork Deep Research](/en-US/features/deep-research).\n\n## The same approach works beyond invoices\n\nThe invoice example follows a simple pattern: bring related materials together, define what matters, and ask for a result someone can act on. That pattern carries over to other work where the answer is spread across several files.\n\n### Compare supplier proposals before making a decision\n\nProvide the proposals, your budget, and the requirements that matter to your team. Ask MuseWork for a side-by-side table covering scope, price, timeline, and exclusions, followed by questions to clarify with each supplier. Separate quoted prices from estimates and cite the page behind each finding.\n\n> Compare these three supplier proposals against the attached requirements. Keep currencies and tax treatment separate. Show differences in scope, price, delivery dates, and exclusions. Cite the source pages, flag missing information, and prepare a decision brief for our project team.\n\n### Turn scattered project materials into a meeting brief\n\nBring together the latest notes, an existing report, and screenshots relevant to the discussion. Ask for a one-page brief that distinguishes decisions already made, unresolved questions, and topics for the meeting. This is especially useful when your colleagues need a concise view of the evidence rather than every original file.\n\n> Prepare a one-page meeting brief from these project materials. Summarize the current situation, list decisions with their sources, flag conflicting or outdated information, and propose an agenda around the unresolved questions.\n\nIn both cases, the materials and the intended reader shape the task. A project team needs a decision brief; meeting participants need the questions worth discussing. Once a first draft is ready, you can refine its depth and format in MuseWork.\n\n## When the source files are already on your computer\n\nFor many of these tasks, the latest files already live in a project folder. MuseWork Desktop can work with local files and data within the access you authorize, and operate the browser with your permission. You can specify the folder to use, the originals to preserve, and where the comparison or brief should be saved.\n\nFor example: “Use the approved documents in this folder, keep the originals unchanged, and save the comparison table and review brief in the output folder.” Keep the desktop client running while the task needs access to your computer.\n\nIf that is where your next task begins, [download MuseWork Desktop](/en-US/download) and start with a small set of files you know well.\n\n## What to know about Terra’s API pricing\n\nIf you are comparing Terra with other models, pricing is another part of the picture. The figures below are OpenAI’s API rates—not MuseWork subscription prices. They help explain the model’s economics, while the document workflows above are where most readers can begin putting it to use.\n\n| Token type | Price per million tokens |\n|---|---:|\n| Input | $1.00 |\n| Cached input | $0.05 |\n| Output | $5.00 |\n\nFor example, **100,000 uncached input tokens and 10,000 billable output tokens cost $0.15 in base token charges**: $0.10 for input plus $0.05 for output. This is a calculation from the rate card, not an estimate of what a particular invoice-review task will use. Billable output can include reasoning tokens.\n\nThe model documentation also lists cache-write charges and different processing rates. Requests exceeding 272,000 input tokens use higher rates for the full request: twice the input and cache rates, and 1.5 times the output rate. Applicable tool fees, retries, and review work also affect the cost of a finished task.\n\n## Frequently asked questions\n\n### Can I use Terra through MuseWork?\n\nYes. We’ve integrated GPT-6.1 Terra into MuseWork. Bring your materials to the workspace and describe the work you want done. If the task depends on files on your computer, the desktop app lets you authorize the access it needs.\n\n### Where can I use GPT-6.1 Terra?\n\nOpenAI’s announcement lists ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, plus the OpenAI API under the model ID `gpt-6.1-terra`.\n\n### Is Terra cheaper than GPT-6.1 Sol?\n\nIts standard input and output token rates are half of Sol’s. The cost of a complete task also depends on reasoning effort, tokens consumed, processing mode, tools, and retries.\n\n### What should I test before giving AI a larger batch of invoices?\n\nStart with a small, approved sample that includes a price mismatch, a missing document, and different tax treatment. Check whether every finding has the correct source and calculation, and whether unknowns stay marked for review. Then decide which steps to extend to more files.\n\n## Sources\n\n- [OpenAI: Introducing GPT-6.1 Terra](https://openai.com/index/introducing-gpt-6-1-terra/)\n- [OpenAI: GPT-6.1 Terra model documentation](https://developers.openai.com/api/docs/models/gpt-6.1-terra)\n- [Artificial Analysis: GPT-6.1 Terra cost-efficiency evaluation](https://artificialanalysis.ai/articles/gpt-6-1-terra-pushes-the-cost-efficiency-frontier)\n- [Simon Willison: GPT-6.1 Terra](https://simonwillison.net/2026/Oct/26/gpt-61-terra/)",[28,31,34,37],{"question":29,"answer":30},"Can I use Terra through MuseWork?","Yes. We’ve integrated GPT-6.1 Terra into MuseWork. Bring your materials to the workspace and describe the work you want done. If the task depends on files on your computer, the desktop app lets you authorize the access it needs.",{"question":32,"answer":33},"Where can I use GPT-6.1 Terra?","OpenAI’s announcement lists ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, plus the OpenAI API under the model ID `gpt-6.1-terra`.",{"question":35,"answer":36},"Is Terra cheaper than GPT-6.1 Sol?","Its standard input and output token rates are half of Sol’s. The cost of a complete task also depends on reasoning effort, tokens consumed, processing mode, tools, and retries.",{"question":38,"answer":39},"What should I test before giving AI a larger batch of invoices?","Start with a small, approved sample that includes a price mismatch, a missing document, and different tax treatment. Check whether every finding has the correct source and calculation, and whether unknowns stay marked for review. Then decide which steps to extend to more files.",1793103556985]