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 what each model is good at, the benchmark 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. All facts come from official OpenAI sources (the 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 AstraとGPT-6.2 Solとは? どちらを使うべき? OpenAIの新モデル2つを、AvaとAndrewの英会話で学びます。最も難しい仕事向けのAstraと、日々の仕事向けで効率的なSol。それぞれの得意分野、ベンチマーク、API料金と注意点、スペック、提供範囲、System cardの要点まで。英日字幕と単語解説つき。ベンチマークの数字はすべてOpenAIの発表値です。公式情報を言い換えたオリジナル対話で、広告ではありません。 ⏱ 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? 3:46 API pricing and the fine print 4:31 Specs: context, output, cutoffs 4:54 Where you can use them 5:35 Safety ratings and safeguards 6:13 Alignment finding + Recap 🔑 Key points / 要点 - Astra: the frontier model for the hardest projects (professional documents, slides, code, science, computer use). Astraは難しい仕事向けです。 - Sol: smarter and more efficient than GPT-6.1 Sol, and close to Astra on some coding tests at much lower cost. Solは日常の仕事向けです。 - 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. 料金の注意点もチェック。 - Both: text and image input, about 1M-token context, up to 128K output tokens, no fine-tuning. 両モデル共通のスペック。 - Safety: OpenAI rates both models Critical for cybersecurity and High for biology/chemistry; Astra is also rated High for AI self-improvement. 安全性の評価も紹介。 - Tip: start with Sol, switch to Astra when a task is really hard, and check that the agent reports problems. 使い分けのコツ。 📚 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 / legitimate / self-improvement / workload / halt / 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-sol/ - OpenAI API docs, GPT-6.2 Astra model 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-6.2 Astra and Sol (what's new): 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 ℹ️ Not sponsored. Independent educational explainer; OpenAI did not review or pay for it. Benchmark results are OpenAI-reported. 本動画はスポンサーなしの教育目的の解説です。 #GPT62 #OpenAI #AIModels #TechEnglish #英語学習 #EigoTechPodcast