GPT-6 Astra Ultra or INK TO BRICK first: Blender cost and time

Conceptual model only. Not a survey or construction document.

Compare INK TO BRICK plus Blender with GPT-6 Astra Ultra through Blender Python or MCP using published prices, cost formulas and workflow stages.

If you already have a floorplan, INK TO BRICK can reconstruct its rooms, walls and openings, let you review the result, and export GLB as the starting point for Blender. GPT-6 Astra Ultra can instead build from a written brief or source plan and operate Blender through its Python API or a separate Blender MCP server. No same-plan benchmark is cited here, so this comparison shows published prices and workflow stages rather than a fixed total or speed multiple.

Note“Ultra” is a useful name for a client preset, but OpenAI’s API documentation names the model gpt-6-astra and its highest documented reasoning effort max. The preset has no separately documented flat API price. API usage is billed by tokens at the model and service tier you choose. See the official GPT-6 Astra model page.

The published prices are different kinds of price

OpenAI publishes higher long-context rates after a request exceeds 272,000 input tokens: Standard becomes $20 input, $2 cached input, $25 cache write and $75 output per million tokens. Batch or Flex costs half the listed Standard rates; Fast costs twice Standard. See the current OpenAI API pricing table before publishing or budgeting a run.

Estimate API token cost

For Standard short-context requests, the API portion is $10 × uncached input millions + $1 × cached input millions + $12.50 × cache-write millions + $50 × output millions. If an illustrative session accumulated 500,000 uncached input tokens and 200,000 output tokens across several requests whose individual inputs stayed at or below the 272,000-token threshold, that arithmetic would be $15. One million cumulative input tokens and 500,000 cumulative output tokens across requests under that threshold would be $35. A single request above 272,000 input tokens uses the long-context rates instead.

A prepaid INK TO BRICK or Codex subscription may already cover the next project within its stated allowances. API tokens remain metered usage even when account credit was purchased earlier. Compare incremental and first-time cost separately: Astra API tokens, Blender Python or MCP setup, retries, compute and review on one side; the INK TO BRICK plan or pass, correction time, export and the Blender refinement you still intend to do on the other. A Codex subscription and OpenAI API billing are separate products.

Where the time goes

The likely practical choice is hybrid. Begin with source-grounded reconstruction in INK TO BRICK, inspect it, export GLB, then let Astra operate Blender through Blender’s Python API or a separate Blender MCP server for custom detail and automation. INK TO BRICK’s own MCP can read existing projects and work with supported Design Option previews; it cannot upload a plan, export a scene or control Blender.

Published Astra architecture example

In OpenAI’s architectural visualization case study, Thomas Ricouard describes Astra building and reviewing a Blender scene through the Blender Python API, then producing a floorplan, interactive home and Unreal Engine tour. Its 30-second figure is the duration of the finished tour.

Worth checkingWhichever route you use, the result is conceptual until checked. A plausible mesh does not establish survey accuracy, structural safety, code compliance or buildability. Use the Blender handoff checks and compare at least two independent dimensions with a trusted source.