OpenAI Launches GPT-6 Astra, an Agent That Operates Computers and Completes Tasks

·1 views

OpenAI this week released GPT-6 Astra, positioning it not as a chatbot upgrade but as an autonomous agent capable of operating browsers, terminals, spreadsheets, and IDEs to complete multi-step tasks. The model features a 1.05 million-token context window and completes OSWorld 2.0 computer-use evaluations roughly 40% faster than its predecessor, GPT-5.6 Sol. Codex task completion is approximately 1.9 times faster on the Mind2Web benchmark, aided by a combined model-and-harness improvement. Pricing is set at around $10 and $50 per million input and output tokens respectively — about 2.5 times the prior model's cost. Rollout is tiered and access-controlled, beginning with enterprise and Trusted Access organizations, with no open weights or self-hosting options available, a decision OpenAI linked to cybersecurity risk concerns.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

How to Build a CI Gate That Catches LLM Output Regressions Before Users Do

A software developer specializing in LLM evaluation has outlined a lightweight, dependency-free method to detect regressions in large language model outputs during the CI/CD pipeline. The approach stores evaluation data as plain JSON files containing questions, expected answers, and match rules, then scores model responses using a single Python script that exits with a non-zero code on failure. Smart matching strategies — including substring checks, numeric tolerance comparisons, and text normalization — reduce false negatives that would otherwise mask real issues. Beyond simple accuracy scores, the method diffs results between two runs to pinpoint which specific test cases regressed, ensuring that an overall score improvement cannot hide breakage in critical queries. For subjective output qualities like tone or format adherence, the author recommends rubric-based model grading at zero temperature rather than string matching alone.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How Composable MCP Servers Can Automate a Solo Developer's Documentation Workflow

Solo developers often struggle to document architectural decisions because their AI tools operate in isolation, forcing manual reconstruction of past work. A composable workflow using two local Model Context Protocol servers aims to solve this by linking session memory with community publishing. The setup pairs a local memory server that indexes past coding sessions via SQLite FTS5 with a DEV.to integration server that scans community discussions and stages draft posts. Together, they allow an AI coding agent to retrieve verified historical benchmarks, identify relevant community gaps, and draft documentation in under five seconds without switching browser tabs. Both servers run locally over stdio, requiring no cloud databases, SaaS subscriptions, or persistent background processes.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

cPanel, Plesk or InterWorx: Which Hosting Control Panel Suits Your Workflow

A detailed technical comparison published on DEV Community evaluates three major web hosting control panels — cPanel, Plesk, and InterWorx — based on real-world use cases rather than feature checklists. All three panels handle core functions such as domains, email, databases, backups, and SSL, but differ significantly in licensing models, supported operating systems, and target users. cPanel is Linux-only, licensed per account, and suits shared-hosting and reseller workflows with the broadest ecosystem of guides and integrations. Plesk supports both Linux and Windows Server with per-server licensing, making it ideal for mixed-OS environments, WordPress fleets, and PostgreSQL or MSSQL stacks. InterWorx, also Linux-only, uses per-server licensing without domain-count restrictions, appealing to cost-sensitive hosting operators who need native multi-server clustering.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Dev.to MCP Server Lets Coding Agents Draft and Sync Technical Articles from Terminal

A developer has released an open-source FastMCP server called dev.to-mcp that connects AI coding agents directly to the DEV.to publishing platform via stdio transport. The tool allows agents to research active community discussions, stage article drafts, and update published posts without the developer leaving their coding environment. By integrating with DEV.to's API, the server exposes functions for searching articles, creating drafts, and syncing post content as the underlying code evolves. The project aims to solve a common problem where technical posts go unwritten or become outdated because writing is disconnected from the development workflow. The complete implementation, including setup instructions and sample prompts, is publicly available on GitHub under the handle kingjulian24.

0 comments Read more at DEV Community