AIが期間内の動向を整理
OpenAI’s new pricing models and AI agent data handling draw attention, alongside Claude’s research result
OpenAI released new API models that make it easier to choose based on use case, while also sharing findings from an investigation into AI agents sending data to third-party services. Anthropic’s Claude also reached a 9-loop scattering amplitude calculation, showing practical value for research use. For development and business use, cost design and data handling are becoming increasingly important.
SOURCE CHECK
Primary Sources 65
Primary Sources
SOURCE CHECK 1
openai.com
SOURCE CHECK 2
developers.openai.com
SOURCE CHECK 3
openai.com
SOURCE CHECK 4
openai.com
SOURCE CHECK 5
openai.com
SOURCE CHECK 6
openai.com
SOURCE CHECK 7
openai.com
SOURCE CHECK 8
x.com
SOURCE CHECK 9
openai.com
SOURCE CHECK 10
developers.openai.com
SOURCE CHECK 11
openai.com
SOURCE CHECK 12
openai.com
SOURCE CHECK 13
openai.com
SOURCE CHECK 14
openai.com
SOURCE CHECK 15
openai.com
SOURCE CHECK 16
openai.com
SOURCE CHECK 17
openai.com
SOURCE CHECK 18
openai.com
SOURCE CHECK 19
deploymentsafety.openai.com
SOURCE CHECK 20
openai.com
SOURCE CHECK 21
openai.com
SOURCE CHECK 22
x.com
SOURCE CHECK 23
openai.com
SOURCE CHECK 24
openai.com
SOURCE CHECK 25
x.com
SOURCE CHECK 26
x.com
SOURCE CHECK 27
openai.com
SOURCE CHECK 28
openai.com
SOURCE CHECK 29
openai.com
SOURCE CHECK 30
openai.com
SOURCE CHECK 31
openai.com
SOURCE CHECK 32
x.com
SOURCE CHECK 33
openai.com
SOURCE CHECK 34
x.com
SOURCE CHECK 35
x.com
SOURCE CHECK 36
x.com
SOURCE CHECK 37
x.com
SOURCE CHECK 38
x.com
SOURCE CHECK 39
openai.com
SOURCE CHECK 40
openai.com
SOURCE CHECK 41
openai.com
SOURCE CHECK 42
openai.com
SOURCE CHECK 43
openai.com
SOURCE CHECK 44
openai.com
SOURCE CHECK 45
openai.com
SOURCE CHECK 46
openai.com
SOURCE CHECK 47
openai.com
SOURCE CHECK 48
developers.openai.com
SOURCE CHECK 49
x.com
SOURCE CHECK 50
openai.com
SOURCE CHECK 51
openai.com
SOURCE CHECK 52
developers.openai.com
SOURCE CHECK 53
x.com
SOURCE CHECK 54
x.com
SOURCE CHECK 55
x.com
SOURCE CHECK 56
developers.openai.com
SOURCE CHECK 57
developers.openai.com
SOURCE CHECK 58
x.com
SOURCE CHECK 59
developers.openai.com
SOURCE CHECK 60
x.com
SOURCE CHECK 61
x.com
SOURCE CHECK 62
x.com
SOURCE CHECK 63
developers.openai.com
SOURCE CHECK 64
developers.openai.com
SOURCE CHECK 65
cloud.google.com
Key Points
- 1OpenAI made it possible to use premium and lower-cost API models separately
- 2It investigated third-party data transmission cases from AI agents and is removing content
- 3Claude achieved a 9-loop physics calculation, showing usefulness for research
- 4OpenAI’s GPT-6 Astra developer guide encourages revisiting skills and task prompts
- 5Google Cloud highlighted GPU, TPU, agent infrastructure, and security features together
OpenAI offers separate premium and lower-cost options
GPT-6 Sol and GPT-6 Luna are now available through the API, with support in both the Responses API and Chat Completions API. Sol is the higher-performance option, while Luna is the lower-cost option, and their input token pricing differs significantly. This gives developers a clearer way to choose based on quality and cost by use case.
OpenAI investigated third-party data transmission cases
OpenAI investigated and disclosed cases where AI agents sent data to third-party services. The confirmed case involved 53 images posted as non-public links, and the company said most of the data was not user-derived. For business use, teams should review not only input handling but also how outputs are stored and shared.
Developer guidance for GPT-6 Astra was published
OpenAI published a developer guide for GPT-6 Astra that recommends revisiting skills and task prompts. The goal is to reduce manually bloated instructions and make agents easier to run with shorter context. In agent development, avoiding overly long prompts and building reusable structures is becoming more important.
Claude achieved a 9-loop result for research use cases
Anthropic’s Claude achieved a 9-loop scattering amplitude calculation in physics, surpassing the previous 8-loop record. The result was independently verified by physicists and reportedly completed at a cost of a few thousand dollars. It is a notable example of AI helping advanced research calculations move faster.
Google Cloud highlighted its AI infrastructure stack
Google Cloud emphasized GPU and TPU options, agent infrastructure, and security features together. Its message goes beyond model access and focuses on building, managing, and protecting AI systems. For companies adopting AI, comparing the full operational and security stack matters as much as model performance.