Samsung's company-wide ChatGPT deployment meets a week of agent infrastructure
Samsung ChatGPT Codex deployment and Claude Tag, and The Deploy Log for this week. Every price and claim read off the publisher's own page, with a call on each.
OpenAI
Samsung ChatGPT Codex deployment
DEPLOYif you run enterprise AI rollouts
WHAT SHIPPEDSamsung Electronics is deploying ChatGPT Enterprise and Codex to all employees in Korea and all Device eXperience division employees worldwide, one of OpenAI's largest enterprise deployments to date. Codex weekly active users in Korea have grown nearly 800% since February 1, 2026, and more than 5 million people now use Codex every week.
THE CATCHThe deployment spans R&D, manufacturing, marketing, and corporate functions, not just software development. Samsung is also supplying memory semiconductors for OpenAI's AI infrastructure, so the relationship covers both workforce adoption and hardware supply.
IF YOU RUNa large organization weighing company-wide AI adoption across technical and non-technical teams
SOURCEOpenAI, Samsung Electronics brings ChatGPT and Codex to employees
Anthropic
Claude Tag
DEPLOYif your team works in Slack
WHAT SHIPPEDClaude Tag is available in beta for Claude Enterprise and Team customers on Slack. Claude joins as a team member, builds context from channels it has access to, plans tasks, works asynchronously, and can schedule tasks for itself over hours or days. Anthropic reports 65% of its product team's code is created by its internal version of Claude Tag.
THE CATCHClaude Tag is multiplayer: one Claude interacts with everyone in a Slack channel, so anyone can see what it is working on and pick up the conversation. Administrators scope access per channel, and memories stay separated between different Claude identities, so a sales setup does not share data with an engineering setup.
IF YOU RUNa team delegating tasks to AI in shared channels with tight access controls
SOURCEAnthropic, Introducing Claude Tag
The Deploy Log
Models and APIs
| DEPLOYif you run document ingestion or RAG pipelines Mistral OCR 4. Mistral OCR 4 supports 170 languages across 10 language groups, returns bounding boxes, block classification, and inline confidence scores, and runs in a single container for self-hosted deployments, priced at $4 per 1,000 pages with a 50% Batch API discount to $2 per 1,000 pages. Mistral |
| DEPLOYif you need lightweight multilingual OCR PP-OCRv6. PP-OCRv6 scales from 1.5M to 34.5M parameters across tiny, small, and medium tiers, with the medium and small tiers supporting 50 languages and the medium tier reaching 86.2% detection Hmean and 83.2% recognition accuracy. Hugging Face |
Tools and products
| DEPLOYif you build TypeScript AI agents AI SDK 7. AI SDK 7, with over 16 million weekly downloads, adds reasoning control, tool and runtime context, provider file and skill uploads, MCP Apps, a terminal UI, tool approvals, WorkflowAgent durability, timeouts, and sandbox support. Vercel |
| DEPLOYif you debug eve agent sessions eve Agent Runs observability. Agent Runs now appears automatically in the Vercel dashboard for every eve project, surfacing trigger, duration, and token usage per session, with retention of 12 hours on Hobby, 1 day on Pro, and 3 days on Enterprise, extendable to 30 days with Observability Plus. Vercel |
| DEPLOYif your team has large repos 500 concurrent builds for Pro. Pro teams can now run up to 500 concurrent builds, up from 12, with on-demand concurrent builds enabled by default for Pro and Enterprise and billing only for build minutes used. Vercel |
Robotics, hardware and chips
| WATCHif you plan large-scale LLM inference infrastructure Jalapeño inference chip. OpenAI and Broadcom unveiled Jalapeño, OpenAI's first Intelligence Processor, co-developed from initial design to manufacturing tape-out in nine months, with engineering samples running ML workloads in the lab at production target frequency and power, including GPT-5.3-Codex-Spark. OpenAI |
Left out
- Gemini 3.5 Flash computer use: the publisher's page does not show it available yet
- GLM 5.2 Fast on AI Gateway: the publisher's page does not show it available yet
- Sakana Fugu Ultra on AI Gateway: the publisher's page does not show it available yet
- Claude Design send-to Vercel: the publisher's page does not show it available yet
- Web Analytics via Vercel CLI: a figure in the draft could not be found on the publisher's own page
- create-chat-sdk scaffolder: a figure in the draft could not be found on the publisher's own page
- Vercel Flags OIDC default: a figure in the draft could not be found on the publisher's own page
- Zero-config Node.js servers: it did not carry a number that moved
- Deep Agents and OpenCode adapters: it did not carry a number that moved
- WebSocket support: it did not carry a number that moved
The pattern
Samsung's company-wide ChatGPT Enterprise and Codex deployment, with Codex weekly active users in Korea up nearly 800% since February 1, 2026 and more than 5 million people using Codex every week, lands in the same week that Anthropic ships Claude Tag in beta for Slack, where 65% of Anthropic's product team code is created by its internal version. The two moves together show enterprise AI shifting from single-user assistants to team-level infrastructure: Samsung spans R&D, manufacturing, marketing, and corporate functions, while Claude Tag is multiplayer, with one Claude interacting with everyone in a channel and administrators scoping access per channel so a sales setup does not share data with an engineering setup.
That shift is reinforced by the tooling and hardware underneath it. Mistral OCR 4 supports 170 languages across 10 language groups at $4 per 1,000 pages with a 50% Batch API discount to $2 per 1,000 pages, and PP-OCRv6 scales from 1.5M to 34.5M parameters with the medium tier reaching 86.2% detection Hmean and 83.2% recognition accuracy, so document intake is now cheap enough to feed agent workflows at Samsung scale. AI SDK 7, with over 16 million weekly downloads, adds WorkflowAgent durability, timeouts, and sandbox support, while Vercel's eve Agent Runs observability surfaces trigger, duration, and token usage per session with retention of 12 hours on Hobby, 1 day on Pro, and 3 days on Enterprise, extendable to 30 days with Observability Plus, and Pro teams can now run up to 500 concurrent builds, up from 12. OpenAI and Broadcom unveiled Jalapeño, OpenAI's first Intelligence Processor, co-developed from initial design to manufacturing tape-out in nine months, with engineering samples running ML workloads in the lab at production target frequency and power, including GPT-5.3-Codex-Spark, and Samsung is supplying memory semiconductors for OpenAI's AI infrastructure, so the same week that agent infrastructure becomes observable and durable, the silicon supply chain for running it at enterprise scale is already in production.