What are GPT-6.2 Astra and GPT-6.2 Sol, and which one should you use? Ava and Andrew explain OpenAI's two new models: Astra, the frontier model for the hardest projects, and Sol, the efficient model for everyday work. They cover results, API prices and the fine print, specs, where you can use them, and the key points from the system card, with English/Japanese subtitles and vocabulary highlights. This is an original, paraphrased educational dialogue, not third-party footage. announcement, the API model pages, the ChatGPT guide and the system card). All benchmark numbers are reported by OpenAI (some come from OpenAI's internal evaluations) and have not been independently verified. Neutral explainer, not an ad. GPT-6.2 어느 쪽을 사용해야합니까? OpenAI의 새로운 모델 2개를 Ava와 Andrew의 영어 대화로 배웁니다. 가장 어려운 업무를 위한 Astra와 일상 업무를 위한 효율적인 Sol. 각각의 전문 분야, 벤치마크, API 요금 및 주의점, 사양, 제공 범위, System card의 요점까지. 영일 자막과 단어 해설 포함. 벤치마크의 숫자는 모두 OpenAI의 발표 값입니다. 공식 정보를 바꿔 말한 오리지널 대화로, 광고가 아닙니다. ⏱ Chapters 0:00 Intro 0:27 Two new models, two jobs 0:54 Astra: professional work and benchmarks 2:09 Computer use and 3D 2:34 Sol: the everyday model 3:18 Which one should you use?3 Specs: context, output, cutoffs 4:54 Where you can use them 5:35 Safety ratings and safeguards 6:13 Alignment finding + Recap 🔑 Key points / 요점- Astra: the frontier model for the hardest projects (professence document, Astra는 어려운 일을위한 것입니다. - Sol: smarter and more efficient than GPT-6.1 Sol, and close to Astra on some coding tests at much lower cost. Sol은 일상 업무용입니다. - API prices per 1M tokens: Astra $8 input / $0.40 cached / $40 output; Sol $2 / $0.10 / $10. Long prompts (over 272K input tokens) and Ultrafast mode cost more; Batch is half price. 요금 주의점도 체크. - Both: text and image input, about 1M-token context, up to 128K output tokens, no fine-tuning. 두 모델 공통의 스펙. - Safety: OpenAI rates both models Critical for cybersecurity and High for biology/chemistry; Astra is also rated High for AI self-improvement. 안전성 평가도 소개. - Tip: start with Sol, switch to Astra when a task is really hard, and check that the agent reports problems. 📚 Vocabulary highlights frontier model / efficient / deliverable / convention / vendor-reported / pass rate / precise / paid plan / merge / pull request / cached input / fine print / context window / knowledge cutoff / fine-tuning / safeguard / legi alignment/obstacle/daily driver/get stuck 🔗 Sources (official, accessed 2026-11-18 JST) - OpenAI, "Introducing GPT-6.2 Astra and Sol": https://openai.com/index/introducing-gpt-6-2-astra-6 page: https://developers.openai.com/api/docs/models/gpt-6.2-astra - OpenAI API docs, GPT-6.2 Sol model page: https://developers.openai.com/api/docs/models/gpt-6.2-sol - ChatGPT Learn, GPT https://learn.chatgpt.com/docs/whats-new/gpt-62-astra-sol - OpenAI, GPT-6.2 Astra and Sol System Card (PDF): https://deploymentsafety.openai.com/gpt-6-2-astra-sol/gpt-6-2-astra-sol.pdf │ OpenAI did not review or pay for it. Benchmark results are OpenAI-reported. #GPT62 #OpenAI #AIModels #TechEnglish #영어 학습 #EigoTechPodcast