GPT-5.6 Launches with Codex Integration and New Productivity Tool ChatGPT Work

OpenAI officially released the GPT-5.6 model family after a limited preview, introducing three variants: the high‑performance Sol model, the work‑balanced Terra model, and the ultra‑fast Luna model. Pricing is tiered at $5 / M input / $30 / M output for Sol, $2.5 / $15 for Terra, and $1 / $6 for Luna.

The series adds a more predictable prompt‑cache mechanism with explicit cache breakpoints, a minimum retention of 30 minutes, write costs at 1.25 × the uncached input rate, and read discounts of 90 % of the input cost.

Benchmark results show Sol achieving a 53.6 score on the Agents’ Last Exam—13.1 points higher than Claude Fable 5—while costing roughly one‑quarter of the estimated expense. Terra and Luna also surpass Fable 5, with costs about one‑sixteenth of its estimate. In coding evaluations, Sol reaches an 80‑point record on the Artificial Analysis Coding Agent Index v1.1, 2.8 points above Fable 5, using less than half the output tokens, under half the runtime, and about one‑third lower cost. Similar advantages appear in Terminal‑Bench 2.1, DeepSWE, and other real‑code workflow tests.

GPT‑5.6’s multi‑agent capabilities enable parallel coordination of four agents by default, improving score‑versus‑latency curves on BrowseComp, SEC‑Bench Pro, and Terminal‑Bench 2.1. Users can access a beta multi‑agent Responses API to replicate the “ultra” experience.

Alongside the model launch, OpenAI merged ChatGPT with Codex to create ChatGPT Work, an agent tool that can directly operate on applications and files, decompose projects into smaller steps, and run for extended periods. It supports Pro, Enterprise, and Edu plans now, with Plus and Business plans arriving soon, and is available on Windows, macOS, and web via the ChatGPT desktop app.

ChatGPT Work can automate tasks such as generating reports from Slack or Notion data, creating marketing briefs, updating presentations from email feedback, and scheduling recurring actions across integrated plugins (e.g., Slack, Teams, Google Drive, CRM). The desktop app also includes a built‑in browser for seamless web‑based workflows.

OpenAI notes that token consumption for ChatGPT Work differs from standard chat requests, scaling with task complexity, and that usage follows the same structure as Codex. The combined release is positioned as a significant shift toward a “super‑app” era, potentially reshaping productivity tools and challenging existing office suites.