README

GPT-6.2 Astra API for Complex Engineering, Science and Professional Work

GPT-6.2 Astra accepts text and images and responds in text across a 1,050,000-token context window, with up to 922,000 input tokens and 128,000 output tokens. It writes cleaner code, builds more capable web frontends and 3D designs, and produces professional deliverables with fewer output tokens.
GPT-6.2 Astra API for Complex Engineering, Science and Professional Work

Key Capabilities of GPT-6.2 Astra

OpenAI positions GPT-6.2 Astra as its flagship model for demanding engineering, scientific research and multistep professional workflows.
Professional Slides, Workbooks and Documents
01

Professional Slides, Workbooks and Documents

OpenAI trained GPT-6.2 Astra to produce clearer slides, spreadsheets and documents that follow the conventions of law, sales, finance and accounting with less filler. Given an original slide file or a screenshot, it matches layout, colors and typography down to the exact font size.
Cleaner Code, Web Frontends and 3D Design
02

Cleaner Code, Web Frontends and 3D Design

GPT-6.2 Astra implements and tests complex changes in existing codebases with cleaner code style and more focused tests. It brings stronger visual reasoning to web design, works on individual parts and connections in CAD tools, and builds interactive 3D browser experiences; on AA-3D, GPT-6.2 Astra ranks first with 1,318 Elo.
1,050,000-Token Context and Up to 128,000 Output Tokens
03

1,050,000-Token Context and Up to 128,000 Output Tokens

Each request supports a 1,050,000-token context window with up to 922,000 input tokens and up to 128,000 output tokens. The model accepts text and image inputs and responds with text, with a May 18, 2026 knowledge cutoff.
Five Reasoning Effort Levels and Pro Mode
04

Five Reasoning Effort Levels and Pro Mode

Guide how much the model thinks with reasoning effort set to low, medium, high, xhigh or max; setting none is rejected with HTTP 400. In the OpenAI Responses API, GPT-6.2 Astra also supports standard and pro reasoning modes.
Scientific Research and Defensive Security
05

Scientific Research and Defensive Security

GPT-6.2 Astra sets new highs on Terminal-Bench-Science 0.1 (+9 percentage points over GPT-6 Astra) and the Vals Science Index for interpreting biological data and running experiments on simulated systems. OpenAI also trained it to write secure code from the start and produce patches that preserve intended behavior.
Responses API Tools and Preserved Reasoning
06

Responses API Tools and Preserved Reasoning

In the OpenAI Responses API, GPT-6.2 Astra supports function calling, web search, file search, structured outputs and subagent delegation. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions to help preserve reasoning across turns and keep earlier input eligible for prompt caching.
SpecificationGPT-6.2 Astra
DeveloperOpenAI
Model ID (OpenAI)gpt-6.2-astra
Release dateNovember 17, 2026
ModalitiesText and image input; text output (audio and video unsupported)
Context window1,050,000 tokens (up to 922,000 input tokens)
Max output tokens128,000 tokens
Reasoning effortlow, medium (recommended starting point), high, xhigh, max; none rejected
Reasoning modesstandard (default), pro (OpenAI Responses API)
Supported featuresStreaming, structured outputs, function calling, web search, file search, image input, prompt caching
Unsupported parameters and featuresParameters: temperature, top_p, logprobs, top_logprobs; features: fine-tuning, predicted outputs
Knowledge cutoffMay 18, 2026
BenchmarkGPT-6.2 Astra ResultWhat It Measures
Agents' Last Exam+5.2 points over GPT-6 Astra (new high)Long-horizon multistep professional workflows
Terminal-Bench-Science 0.1+9 percentage points over GPT-6 Astra (new high)Scientific work carried out in a terminal
GDPval v2+6.2 percentage points over GPT-6 Astra with 22% fewer output tokensMultistep consulting, finance and design assignments
AA-AnalystAgent62.5% of questions answered correctly on all 5 independent attempts (new high)Financial and expenditure analysis over real spreadsheets and reports
AA-3D1,318 Elo, ranked first (GPT-6.2 Sol: 1,259 Elo)Blinded human preference on interactive 3D browser builds
BenchCAD 2.1 Core32% improvement over GPT-6 AstraBuilding 3D parts and assemblies from engineering drawings
Cybersecurity evaluationsSystem Card targets: 36.4% solved vs 18.2% for GPT-6 Astra; ExploitBench Internal Port: +22 points over GPT-6 AstraSeparate security evaluations reported by OpenAI

How to Work With GPT-6.2 Astra

Four OpenAI API practices that govern accuracy, tool use and token efficiency.
  1. 1

    Step 1: Choose Between GPT-6.2 Astra and GPT-6.2 Sol

    Use GPT-6.2 Astra for your hardest projects across engineering, scientific research, professional deliverables, frontends and 3D design. For day-to-day coding, OpenAI recommends starting with GPT-6.2 Sol, its efficient, versatile model for everyday work.
  2. 2

    Step 2: Set Reasoning Effort From Low to Max

    OpenAI recommends starting with medium for a new integration, or switching legacy none and minimal configurations to low; setting reasoning effort to none returns HTTP 400. Supported effort levels are low, medium, high, xhigh and max. Omit temperature, top_p, logprobs and top_logprobs from your requests.
  3. 3

    Step 3: Keep Prompt Prefixes Unchanged

    Starting with GPT-6.2, OpenAI's API drops reasoning items when it detects mid-conversation prefix edits (such as changes to instructions or tool calls), a model downgrade, or a change in the end user or organization. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions to help preserve reasoning across turns and keep earlier input eligible for prompt caching.
  4. 4

    Step 4: Use the Responses API for Tool Calling

    GPT-6.2 Astra supports both Chat Completions and the Responses API, but tool calling requires the Responses API. Because GPT-6.2 Astra writes more concisely and directly by default, review legacy system prompts that repeat "be concise" instructions.

What You Can Build With GPT-6.2 Astra

Typical workloads for OpenAI's 1.05M-context frontier model.

Complex Codebase Changes and Security Patches

Implement and test multi-file changes across large repositories, write secure code from the start and generate patches that preserve intended behavior.

Financial Workbooks, Legal Briefs and Slide Decks

Analyze figures across spreadsheets and expenditure reports, find facts in legal source documents and generate slides that match your existing templates.

Interactive Web Frontends and 3D Browser Apps

Turn written briefs and visual references into distinctive web interfaces, browser games, CAD mechanical models and interactive 3D scenes.

Scientific Computing and Data Interpretation

Run terminal-based scientific workflows, write Python solvers for physical simulations and interpret biological or experimental datasets.

GPT-6.2 Astra on Kie.ai

01

GPT-6.2 Astra API on Kie.ai

Kie.ai is adding GPT-6.2 Astra to its chat models. This page is the reference for the documented capabilities, specifications and benchmarks of the model.
02

Chat Models in the Kie.ai Market

Browse the chat models in the Kie.ai market, including GPT-6 Astra, GPT-6.1 Sol and GPT-6.1 Terra, alongside the GPT-6.2 Sol overview.
03

One Key Across the Kie.ai Market

All models in the Kie.ai market use the same authentication method: an Authorization: Bearer YOUR_API_KEY header with the key from your Kie.ai API Key Management page.
04

One Credit Balance

Your account has one credit balance that all models in the Kie.ai market draw from, and you can check it with the Get Remaining Credits API.

Frequently Asked Questions

What is GPT-6.2 Astra?

GPT-6.2 Astra (gpt-6.2-astra) is OpenAI's frontier model for demanding engineering, scientific research and complex professional deliverables, released on November 17, 2026 alongside GPT-6.2 Sol.

How is GPT-6.2 Astra different from GPT-6.2 Sol?

Both models share a 1,050,000-token context window, a 128,000-token output limit, and support for streaming, structured outputs, function calling, web search, file search, image input and prompt caching. OpenAI positions GPT-6.2 Astra as its flagship model for the hardest projects, and GPT-6.2 Sol as its efficient, versatile go-to model for day-to-day coding and everyday work.

What are the context window and token limits of GPT-6.2 Astra?

GPT-6.2 Astra has a 1,050,000-token context window, accepts up to 922,000 input tokens and generates up to 128,000 output tokens. Its knowledge cutoff is May 18, 2026.

Which reasoning effort levels does GPT-6.2 Astra support?

It supports low, medium, high, xhigh and max via reasoning.effort in the Responses API or reasoning_effort in Chat Completions. Requests that set reasoning effort to none are rejected with HTTP 400; OpenAI recommends medium for new integrations and low when migrating from none or minimal.

Does GPT-6.2 Astra support Chat Completions and tool calling?

GPT-6.2 Astra supports both Chat Completions (v1/chat/completions) and the Responses API (v1/responses), as well as Batch (v1/batch) in OpenAI's API, but tool calling requires the Responses API.

Why does OpenAI recommend keeping prompt prefixes unchanged?

Starting with GPT-6.2, OpenAI's API drops reasoning items when it detects mid-conversation prefix edits (such as changes to instructions or tool calls), a model downgrade, or a change in the end user or organization. OpenAI recommends keeping prompt prefixes unchanged and appending new instructions to help preserve reasoning across turns and keep earlier input eligible for prompt caching.

What inputs and sampling parameters are unsupported in GPT-6.2 Astra?

GPT-6.2 Astra accepts text and image inputs and outputs text; audio and video modalities, fine-tuning and predicted outputs are unsupported. Omit temperature, top_p, logprobs and top_logprobs when sending requests.

Is the GPT-6.2 Astra API available on Kie.ai?

Not yet. Kie.ai is adding GPT-6.2 Astra to its chat models. This page covers OpenAI's documented capabilities, parameters and benchmarks for the model.