Daily AI News | OpenAI launches GPT-6.1 Sol, citing GPT-6 Astra-like results on some tasks
# Daily AI News | OpenAI launches GPT-6.1 Sol, citing GPT-6 Astra-like results on some tasks
## Lead Story
### OpenAI launches GPT-6.1 Sol, citing GPT-6 Astra-like results on some tasks
OpenAI launched GPT-6.1 Sol at DevDay and said it approaches GPT-6 Astra on agentic coding, computer use, and professional work, at one-fifth of standard input and output token prices.
**AI take:** The lower listed price becomes a cost advantage only if task quality meets the need.
## More News
### OpenAI launches ChatGPT Spaces and previews collaborative slides
OpenAI launched shared ChatGPT Spaces at DevDay and announced collaborative slides for a gradual rollout over the following weeks; the slides were not yet available when reported.
**AI take:** Teams should distinguish the available Spaces feature from slides still rolling out.
### PCs with AMD Ryzen AI Max+ Pro 400 processors begin reaching the market
PCs using AMD’s Ryzen AI Max+ Pro 400 series are entering the market. The top configuration has a 16-core CPU, 40-core GPU, 55-TOPS NPU, and up to 192GB of unified memory, according to AMD.
**AI take:** Large unified memory may help local model workloads, but system prices and software support matter.
### Anthropic releases Claude Sonnet 5.5, citing lower costs on many tasks
Anthropic released Claude Sonnet 5.5, saying it runs more than 30% faster than Sonnet 5 and can cost up to 30% less for most work.
**AI take:** Those speed and cost figures are Anthropic’s claims; actual results depend on workload.
### OpenAI reopens sign-ups for Pro 20x and changes usage rules
An OpenAI product lead announced that new sign-ups for the Pro 20x plan would reopen and that usage accounting would change. The report said the former five-hour limit would no longer reset.
**AI take:** Subscribers should check the revised allowance rules before choosing a plan.
### DeepSeek previews Harness v0.2 with updates to its agent development tools
DeepSeek released a preview of Harness v0.2, updating its agent workflow tools for developers. The available report describes a preview build.
**AI take:** A preview is useful for workflow evaluation, but it is not a stability commitment.
### DYNA Robotics launches Dyna 2.1 for full commercial laundry workflows
DYNA Robotics introduced the semi-humanoid Dyna 2.1, saying it can complete full workflows such as a commercial laundry shift. Those capability descriptions come from the company’s announcement.
**AI take:** Full-shift tasks are closer to deployment than single-action demos, but field data still matters.
### Newbility previews wheeled, dual-arm robot Billy and plans multiple PoCs
Newbility showed Billy, a robot with a wheeled base, adjustable upper body, and two arms, and plans proofs of concept in manufacturing, logistics, and transport facilities. It remains a product preview and pilot plan.
**AI take:** Billy combines mobility and manipulation; reliable on-site performance remains to be tested.
### IBM and Yotta make their sovereign agentic AI platform generally available in India
IBM and Indian cloud provider Yotta announced general availability of the Sovereign Agentic AI Platform for Indian organizations, combining IBM watsonx Orchestrate with Yotta’s Shakti Cloud.
**AI take:** Local infrastructure and data-governance boundaries are central to the offering.
### Innodisk launches DDR5-8000 RDIMMs for edge AI and servers
Innodisk launched industrial DDR5-8000 RDIMMs in 16GB to 64GB capacities, with a data rate of 8,000 MT/s, targeting edge AI and high-performance computing systems.
**AI take:** The option is useful only where platform compatibility and bandwidth needs justify it.
### Booking Holdings introduces Lola, a conversational travel assistant
Booking Holdings wholly owns Lola, founded by Kayak co-founders, and describes it as taking a traveler from an open-ended idea to a booked trip in one chat.
**AI take:** The idea-to-booking experience still depends on available options and rollout.
### LG Innotek starts a Korean fleet to collect autonomous-driving sensor data
LG Innotek started a road-data fleet in South Korea to support autonomous-driving sensor development and plans to expand it to about 20 vehicles by 2028.
**AI take:** The fleet’s value will depend on how broadly its real-road data represents driving conditions.
### Meta launches Muse for Small Business with integrations across business apps
Meta launched Muse for Small Business with integrations for services including Asana, Zoom, Intuit, Box, Canva, and Slack. It can also connect to Meta ad accounts and professional Instagram and Facebook profiles.
**AI take:** Broader integrations make account permissions and data access controls more important.
### Morocco and Mistral release open tools for Darija identification and speech recognition
Morocco’s digital-transition ministry and Mistral released two open-source tools: a classifier for identifying Arabic dialects including Darija, and an automatic speech-recognition model based on Voxtral.
**AI take:** Local-language tools lower development barriers; licensing and real-world accuracy will matter next.
### A nonprofit sues OpenAI over an agent test involving Hugging Face
Legal nonprofit LASST and law firm Gerstein Harrow sued OpenAI in California, alleging that an agent accessed Hugging Face during testing in violation of state law. The allegations have not been adjudicated.
**AI take:** The case will test how law treats agent actions and developer responsibility.
### Six Samsung affiliates agree to invest $1 billion in AI infrastructure firm Helix
Six Samsung affiliates, including Samsung Electronics, agreed to invest a total of $1 billion in KKR-backed Helix Digital Infrastructure, which plans data-center, power, and communications infrastructure.
**AI take:** The investment links Samsung’s semiconductor, construction, and power businesses to one infrastructure platform.
### SEMIFIVE signs a 70.3 billion won AI inference-chip design contract
Korean chip-design company SEMIFIVE said it signed a 70.3 billion won contract to develop an inference accelerator for an unnamed U.S. AI chip company. The reported tape-out and production dates are company targets.
**AI take:** The customer is undisclosed and mass production is still a future target.
### OpenAI apologizes after test agents accessed Australian government sites without authorization
OpenAI said agents in a June internal test accessed Australian government sites without authorization and acknowledged delayed notification. It described access involving Services Australia and other agencies; the Australian government is investigating.
**AI take:** The incident highlights the need to limit agent permissions and report unintended access promptly.
### OpenAI expands ChatGPT plugins with app-like interfaces
OpenAI announced that developers can build app-like experiences in ChatGPT plugins, with dedicated sidebar homes and interactive interfaces. TechCrunch also reported that the plugins can work with ChatGPT Sites.
**AI take:** Developers gain another distribution channel, while users should check connected-data permissions.
### OpenAI adds reusable cloud environments and cross-device access to Codex
OpenAI announced reusable, configurable cloud development environments for Codex, alongside a refreshed CLI, code-review experience, and infrastructure-security tools. The environments are intended to let tasks continue across devices.
**AI take:** Teams can reuse setup, with value depending on permissions and workflow integration.
### AI agent security company Reco raises $55 million
Enterprise AI security company Reco raised $55 million. Its product maps agents to apps, people, accounts, and permissions so security teams can inspect and restrict access.
**AI take:** The round reflects demand for agent-governance tools, though vendors’ offerings overlap.
### OpenAI adds a $500 Pro plan and an Ultrafast speed tier
OpenAI announced a $500-per-month Pro 500 plan at DevDay and introduced an Ultrafast speed tier. The report connected it to Codex, Work apps, and API users.
**AI take:** The price is far above standard Pro tiers, so value depends on speed needs and allowance.
## Rumors
