Model · OpenAI

GPT-6 Astra

Available with every Share plan

Run GPT-6 Astra in an Agent.Space Workspace

Pair OpenAI’s high-capability model with a compatible coding Agent. Agent.Space keeps the model choice, project files, Sessions, previews, and review context in one persistent cloud Workspace.

Open Agent.SpaceCompare plans
GPT-6 Astra × Agent.Space
  1. 01
    Agent harness

    Choose the execution workflow separately

  2. 02
    GPT-6 Astra

    Select the model and reasoning level

  3. 03
    Workspace

    Keep files, Sessions, and results together

Official model capacity

context window
1.05M
maximum output
128K
reasoning effort
Low–Max

01 · Model or Agent?

A powerful model still needs an execution layer

GPT-6 Astra supplies the reasoning. The Agent harness decides how to inspect a project, call tools, edit files, and verify a result. Agent.Space keeps those choices explicit.

MODEL

GPT-6 Astra is the model

It reads the selected context and produces reasoning, text, code, and tool decisions. OpenAI’s API documentation defines its model-level capabilities and limits.

HARNESS

The Agent harness runs the task

Codex, Claude Code, OpenCode, and other harnesses each shape a different execution loop. Choose a compatible combination in the live Workspace instead of inferring it from the model name.

02 · Task fit

Use Astra when the task can earn the extra capability

Start with the cost of a wrong answer, the amount of connected evidence, and the difficulty of verification—not with the newest model label.

Strong candidate for

  • Ambiguous, high-value changes that cross many related files.
  • Long investigations that connect code, logs, screenshots, and documents.
  • Work where fewer failed turns may matter more than the lowest token price.

Compare a lighter model for

  • Routine formatting, extraction, or deterministic transformations.
  • High-volume repetitive tasks with simple acceptance checks.
  • Small fixes where a narrower context and faster response are enough.

03 · Agent.Space workflow

Give the model a Workspace you can return to

The model is one choice inside a longer workflow. Agent.Space preserves the project surface around the run so people and other Agents can inspect or continue the work.
01

Persistent project context

Keep project files and Sessions available beyond one local terminal window.

02

Explicit model controls

Select the Agent harness, compatible model, and available reasoning level as separate choices.

03

Review and handoff

Inspect files, previews, and results before a teammate or another Agent continues.

Agent.Space cloud Workspace showing project navigation, Sessions, and the task composer
A persistent Workspace keeps the project and review surface around the model run. The live selector remains the source of truth for compatible Agents and controls.
  1. 01

    Open the project Workspace

    Start from the files and environment the task must change.

  2. 02

    Choose a compatible Agent

    Pick the harness whose execution style fits the work.

  3. 03

    Select GPT-6 Astra

    Choose the model and an available reasoning level deliberately.

  4. 04

    Run, inspect, and continue

    Review the result in the same Workspace before handoff.

04 · Access and boundaries

Available in Share, governed by the live selector

Agent.Space currently includes GPT-6 Astra in every Share plan. The pricing page defines the current commercial terms; the Workspace defines the combinations available to your account.
SHARE × GPT-6 ASTRA

GPT-6 Astra in every Share plan

Choose a Share plan, open a Workspace, then select Astra where the chosen Agent supports it. Exact allowance and plan terms stay on the pricing page so this page does not duplicate changing billing details.

View current plans

Check before a real task

  1. 01

    Model availability and compatible Agent combinations come from the live Workspace selector.

  2. 02

    OpenAI model-level tools or modalities are not automatically implemented by every Agent harness.

  3. 03

    A larger context window does not replace scoped instructions, permissions, verification, or human review.

05 · Next step

Choose with current product and source information

Agent.Space plans

Review current Share access, allowances, and billing terms.

Open resource

Workspace overview

See how files, Sessions, previews, and handoff stay together.

Open resource

Official model specifications

Check OpenAI’s latest context, output, pricing, and capability details.

Open resource

Agent.Space: a GPT-6 Astra access alternative for coding

Use GPT-6 Astra in Agent.Space from $9/month. Get a compatible coding Agent, persistent cloud Workspace, project files, Previews, and checks—no local setup.

Open resource

AI code review that checks the project, not just the diff

Review a code diff with GPT-6 Astra in a compatible coding Agent, then verify findings against project files, checks, and human acceptance criteria.

Open resource

GPT-6 Astra game generator case: a playable browser game

See Checkmate Bastion, an original playable browser game case, and the reviewable workflow for building a game with GPT-6 Astra and a coding Agent.

Open resource

06 · GPT-6 Astra FAQ

What to know before you select it

Short answers about Agent.Space access, Agent compatibility, and task choice.

Start in one Space

Put GPT-6 Astra behind a task you can verify

Open a persistent Agent.Space Workspace, choose a compatible Agent and model combination, and begin with one bounded result.

Open Agent.Space