OpenAI × New Model Releases

OpenAI recaps its DevDay 2026 launches

OpenAI recaps its DevDay 2026 launches

✎ Story body

OpenAI recapped DevDay 2026, while ChatGPT plan changes and an AWS agent toolkit moved in the same period.

What happened

OpenAI published its DevDay 2026 recap. Japanese outlets reported that ChatGPT's top plan reopens to new sign-ups with the 5-hour limit removed, and that an "Ultrafast" mode is tied to the $500/month tier. AWS open-sourced its Strands harness for building AI agents.

Why it matters

Developer announcements and plan or pricing changes landed in the same week, shifting usage terms noticeably. AWS's model-agnostic open-source release is a separate thread, but it reads alongside OpenAI's moves as agent-platform competition.

What to watch

Watch when DevDay features reach general availability, how the top-tier quotas and pricing work in practice, and how widely the Strands harness is adopted.

▲ Official & Press
Official

DevDay 2026 Recap

OpenAI Blog ・ 2026-09-29 ・ 📌

OpenAI recaps 20+ DevDay 2026 announcements, incl. GPT-6 Astra

Press

OpenAIの「DevDay 2026」で発表された主なことまとめ

ITmedia AI+ ・ 2026-09-29

OpenAI DevDay 2026: 20+ launches incl. always-on agent dots, GPT-6.1 Sol

Press

「GPT-6.1 Sol」登場 「一部Astra並み」性能をコスト5分の1で提供 先代モデルからわずか1週間で公開

ITmedia AI+ ・ 2026-09-29

OpenAI launches GPT-6.1 Sol: near-Astra performance at one-fifth the price

Press

ChatGPTに「Ultrafast」、Astraを8倍高速に 利用には「月500ドル」最上位プランが必須

ITmedia AI+ ・ 2026-09-29

OpenAI adds 'Ultrafast' mode running GPT-6 Astra up to 8x faster

Press

OpenAI、「24時間働く」エージェント「dots」発表 Astra搭載、クラウド上で常時稼働

ITmedia AI+ ・ 2026-09-29

OpenAI unveils 'dots', an always-on agent powered by GPT-6 Astra

Community

GPT 6.1 Sol

Hacker News (Front Page) ・ 2026-09-29

OpenAI publishes an introduction page for "GPT 6.1 Sol"

Press

ChatGPT、サブスク最上位プランの新規受付を再開へ “5時間制限”を完全撤廃 「利用枠を消費しない新機能」も

ITmedia AI+ ・ 2026-09-29

OpenAI reopens ChatGPT Pro (20x) signups, drops 5-hour limit

Press

AWS、AIエージェントを自作できるツール「Strandsハーネス」をオープンソースで公開 特定のLLMに依存せず入れ替え可能

ITmedia AI+ ・ 2026-09-29

AWS open-sources Strands harness for building LLM-agnostic AI agents

Press

AWS、AIエージェントを自作できるツール「Strandsハーネス」をオープンソースで公開。特定のLLMに依存せず入れ替え可能、任意のコンテナ環境にデプロイ

Publickey ・ 2026-09-28

AWS open-sources Strands, a multi-LLM AI agent harness

Academic (arxiv etc.) 13 ▾
Academic

Rethinking Circuit Evaluation: Do Circuits Explain Model Errors?

arXiv cs.LG (Machine Learning) ・ 2026-09-28

Academic

Not All Thinking is Created Equal: Latent Reasoning Discovers a Recurrent Search Algorithm for Depth Generalization

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

Verifiable Visual Rewards Transfer from Synthetic Scenes to Natural Prompts

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

TCSAlgBench: Benchmarking Automated Proving for Research-Level Theoretical Computer Science

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

Share-Borne AI Virus: Memory-Hopping Attacks Across LLM Agents

arXiv cs.LG (Machine Learning) ・ 2026-09-28

Academic

Almieyar: A Culturally Grounded Benchmark for Multi-Dialect Arabic Speech Recognition

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

RareDx: Controlled Knowledge Integration and Graph-Grounded Policy Optimization for Rare-Disease Diagnosis

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

Epistemic Policy Divergence in Multi-Turn LLM Contamination: A Protocol-Gradient Investigation

arXiv cs.CL (Computation and Language) ・ 2026-09-28

Academic

Decide, Don't Generate: Competitive Dimensional ABSA with Jev's Typed Decisions

arXiv cs.CL (Computation and Language) ・ 2026-09-28

Academic

Timeline-Bench: Evaluating Agents on Realistic Video-Editing Tasks, from Raw Footage to Final Cut

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

What Drives Citations in Production Large Language Models? An Observational Multi-Method Study of Two Million AI Citations Across Ten Thousand Web Pages

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

APEX-Voice: Can Voice Agents Complete Professional Workflows Through Full-Duplex Interaction

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

Academic

JevVibe: Efficient Classification-Guided Secure Code Generation

arXiv cs.AI (Artificial Intelligence) ・ 2026-09-28

← Story Archive