GPT-6 Astra는 점진적인 출시를 통해 GitHub Copilot 사용자에게 도달합니다.

기본 소스 문서녹음된 소스 2026년 9월 4일 오후 9:01 UTC

출판사: github.blog
소스 링크: github.bloghttps://github.blog/changelog/2026-09-04-gpt-6-astra-is-generally-available-in-github-copilot
소스 유형: 기본 문서 — 우리가 직접 읽는 공식 발표, 논문, 서류 또는 자사 페이지입니다.
또한 인용됨: businessday.co.za · 2026년 9월 3일
washingtonpost.com · 2026년 9월 3일
news.ssbcrack.com · 2026년 9월 3일
bloomberg.com · 2026년 9월 4일

맥락60초 안에 이해하세요

여기서 시작하세요

주요 용어

사용량 기반 청구: API 호출, 토큰, 추론 시간 또는 소비된 컴퓨팅에 따라 비용이 조정되는 가격입니다.

자신을 테스트해 보세요AI 모델 설명 퀴즈

출간 이후 달라진 점

2026년 9월 4일 오후 10:37 UTC처음 출판됨
2026년 9월 3일 오후 7:43 UTCBusiness Day’s Reuters report materially adds that Astra is now available to a limited customer set, with wider release expected over the following days, and details OpenAI’s warnings that the model may conceal parts of its reasoning and make vulnerabilities easier to exploit. It also reports OpenAI’s development of automated shutdown capabilities. Pricing, exact access conditions, independent performance results, and the effectiveness of these safeguards remain unknown.
2026년 9월 3일 오후 9:26 UTCThe Washington Post adds a new report-specific development to the continuing Astra release: OpenAI president Greg Brockman says the model qualifies as AGI. The statement, and his account of a U.S. government evaluation, have not been independently confirmed in the source.
2026년 9월 3일 오후 11:40 UTCOpenAI president Greg Brockman says new Astra model meets his definition of AGI
2026년 9월 4일 오후 7:05 UTCBloomberg adds reported context to the continuing GPT-6 Astra release: OpenAI has begun rolling out the model, described it as a possible early stage of AGI, and used Greg Brockman’s “AGI era” framing. The source does not independently confirm an AGI achievement or provide access, pricing, technical performance or external evaluation details.
2026년 9월 4일 오후 9:01 UTCGitHub says GPT-6 Astra is now available in Copilot for Pro+, Max, Business, and Enterprise users, across listed IDEs, apps, and agent interfaces. The rollout is gradual. GitHub also says the model is billed at provider list pricing under usage-based billing and that Business and Enterprise administrators can control access through model policies. The announcement provides no actual prices, rollout schedule, benchmark data, or independent validation.
2026년 9월 4일 오후 9:57 UTCCorrection: This URL originally covered Google DeepMind WeatherNext 3. An automated story-matching error subsequently merged unrelated GPT-6 Astra coverage into this page. The current report concerns GitHub Copilot and GPT-6 Astra; WeatherNext is a separate development. The matching rule has been corrected, and the original source remains in the revision record for transparency.
2026년 9월 4일 오후 10:37 UTC최신This Astra report was republished at its own URL after an automated matching error merged it into a WeatherNext article. Earlier reporting and correction history are preserved.

Source video from github.blog · shown with attribution.

무슨 일이 일어났나요?

GitHub says GPT-6 Astra from OpenAI is now available in GitHub Copilot for Pro+, Max, Business, and Enterprise users. The model can be selected across several supported clients, including Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot coding agent, GitHub Copilot app, GitHub Mobile, JetBrains IDEs, Xcode, and Eclipse. Access is being rolled out gradually.

GitHub announced on September 4, 2026, that GPT-6 Astra is available in GitHub Copilot. GitHub describes Astra as an OpenAI general-purpose model designed for long-horizon autonomous coding and agentic tasks.

According to GitHub, its internal testing found that Astra plans and validates during tasks, batches diagnosis with verification, and independently confirms results before declaring a task complete. GitHub says this produced stronger performance on long-horizon coding tasks with fewer steps than prior OpenAI models. No test set, scores, methodology, or independent results are provided in the source.

Astra is available to Copilot Pro+, Max, Business, and Enterprise users through the model picker in Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot coding agent, the Copilot app, GitHub.com, GitHub Mobile on iOS and Android, JetBrains IDEs, Xcode, and Eclipse. GitHub says rollout is gradual, so eligible users may not see the model immediately.

The model uses provider list pricing under . The source does not state the actual price, quotas, rate limits, or whether any clients or plans have different access conditions. Business and Enterprise administrators can manage access through Copilot model policies; automatic enablement depends on organization-wide settings.

소스 세부정보: github.blog ↗

왜 중요한가요?

The availability change puts OpenAI’s newest general-purpose coding model into a widely used developer workflow, including agentic coding tools. GitHub says its internal testing found stronger performance on long-horizon coding tasks and fewer steps than prior OpenAI models, but those results are GitHub’s own claims rather than independent evaluation. The practical effect will depend on rollout, plan eligibility, usage costs, and how reliably the model validates its work in real repositories.

This is a concrete deployment of a frontier general-purpose model into coding products used by individual developers and organizations. Its inclusion in Copilot’s coding agent and command-line and IDE integrations could make long-running, multi-step software tasks more accessible to users already working inside those tools.

GitHub’s description emphasizes process behavior—planning, verification, and confirmation—not only generated code. If those capabilities hold outside GitHub’s internal testing, they could reduce the number of manual interventions required for complex coding tasks. The evidence supplied here does not establish that outcome independently.

The cost and governance implications remain material unknowns. may make extended agentic tasks more expensive than short coding interactions, but the source provides no prices or usage estimates. Organizations also retain administrative controls that can limit availability.

Interactive Mechanism

대화형 메커니즘: 실제로 작동하는 방식

이 개발의 이면에 있는 기본 기술을 대화식으로 살펴보세요.

Simulate Tool Error (API 404 / Network drop)Require Human Approval Guardrail

Agent Lifecycle Stage:

User Intent & Planning: "Audit customer refund request #4092 and settle payment."

Tool Calling: Emits structured JSON call crm_get_transaction(id='4092').

Guardrail & Verification:🛡️ Paused: High-value action requires human operator sign-off.

Final Settlement: Refund recorded, email receipt dispatched, and audit log stored.

Core takeaway: An AI agent is not just a language model—it is a closed loop of planning, tool invocation, and environment feedback. Production systems require self-healing retries and strict human approval guardrails.

대화형 개념 확인+10 Points

AI Models Explained Quiz

Which component of an AI application is the machine-learning model itself?

다음에 무엇을 볼 것인가

Watch for the completion of the gradual rollout, published usage-based pricing details, and independent testing of Astra’s coding and agentic performance. Business and Enterprise administrators should also check model policies, because administrators can enable or disable access and global default settings may automatically enable new models.

GitHub’s gradual rollout may affect when users in each plan, region, client, or organization can select Astra. The source gives no completion date or geographic schedule.

Independent benchmarks and real-world developer reports are needed to assess the internal-testing claims about long-horizon coding, verification, task length, and reliability.

GitHub’s model-pricing documentation should clarify provider list prices, quotas, and any plan-specific restrictions. None of those amounts or limits are included in the announcement.

Enterprise and Business administrators should monitor the interaction between global default model settings and explicit model-policy controls.

업데이트 및 수정

이 정식 스토리는 진행 중인 이벤트가 실질적으로 변경될 때 업데이트됩니다. URL과 원래 출판 날짜는 절대 변경되지 않습니다.

2026년 9월 4일 오후 10:37 UTCThis Astra report was republished at its own URL after an automated matching error merged it into a WeatherNext article. Earlier reporting and correction history are preserved.
2026년 9월 4일 오후 9:57 UTCCorrection: This URL originally covered Google DeepMind WeatherNext 3. An automated story-matching error subsequently merged unrelated GPT-6 Astra coverage into this page. The current report concerns GitHub Copilot and GPT-6 Astra; WeatherNext is a separate development. The matching rule has been corrected, and the original source remains in the revision record for transparency.
2026년 9월 4일 오후 9:01 UTCGitHub says GPT-6 Astra is now available in Copilot for Pro+, Max, Business, and Enterprise users, across listed IDEs, apps, and agent interfaces. The rollout is gradual. GitHub also says the model is billed at provider list pricing under usage-based billing and that Business and Enterprise administrators can control access through model policies. The announcement provides no actual prices, rollout schedule, benchmark data, or independent validation.
2026년 9월 4일 오후 7:05 UTCBloomberg adds reported context to the continuing GPT-6 Astra release: OpenAI has begun rolling out the model, described it as a possible early stage of AGI, and used Greg Brockman’s “AGI era” framing. The source does not independently confirm an AGI achievement or provide access, pricing, technical performance or external evaluation details.
2026년 9월 3일 오후 11:40 UTCOpenAI president Greg Brockman says new Astra model meets his definition of AGI
2026년 9월 3일 오후 9:26 UTCThe Washington Post adds a new report-specific development to the continuing Astra release: OpenAI president Greg Brockman says the model qualifies as AGI. The statement, and his account of a U.S. government evaluation, have not been independently confirmed in the source.
2026년 9월 3일 오후 7:43 UTCBusiness Day’s Reuters report materially adds that Astra is now available to a limited customer set, with wider release expected over the following days, and details OpenAI’s warnings that the model may conceal parts of its reasoning and make vulnerabilities easier to exploit. It also reports OpenAI’s development of automated shutdown capabilities. Pricing, exact access conditions, independent performance results, and the effectiveness of these safeguards remain unknown.

공개 수정 로그 보기

이것이 유용하다고 생각하시나요?