OpenAI’s GPT-5 is live—and Microsoft has switched on the new model across its Copilot stack, from Microsoft 365 Copilot and the consumer Copilot to GitHub Copilot and Azure AI Foundry, with Perplexity also lighting it up for its Max and Pro tiers (including the Comet AI browser). (openai.com, techcommunity.microsoft.com, github.blog, azure.microsoft.com, thetechoutlook.com)
OpenAI describes GPT-5 as its “smartest, fastest, and most useful” model to date, available in ChatGPT immediately, with Team access now and Enterprise/Edu rollout dated August 14. The public product page also lists three API sizes—GPT‑5, GPT‑5 mini, and GPT‑5 nano—plus a 400K context window, new “minimal reasoning” and “verbosity” controls, and updated pricing. (openai.com)
Source: The Tech Outlook ChatGPT-5 launches across Microsoft 365 Copilot, Copilot, GitHub Copilot and Azure AI Foundry; Also available on Perplexity and Comet for Max and Pro users - The Tech Outlook
Overview
OpenAI describes GPT-5 as its “smartest, fastest, and most useful” model to date, available in ChatGPT immediately, with Team access now and Enterprise/Edu rollout dated August 14. The public product page also lists three API sizes—GPT‑5, GPT‑5 mini, and GPT‑5 nano—plus a 400K context window, new “minimal reasoning” and “verbosity” controls, and updated pricing. (openai.com)- GPT‑5 (text+vision): 400K context, up to 128K output tokens, $1.25/M input tokens, $10/M output tokens. (openai.com)
- Positioning: stronger at coding, reasoning, safer responses, and health answers; “thinks harder” on complex requests. (openai.com)
What’s new across Microsoft’s Copilot family
Microsoft 365 Copilot and Copilot Studio
Microsoft says GPT‑5 is rolling out now in Microsoft 365 Copilot and Copilot Studio worldwide. Expect better multi-step reasoning on long threads, tighter context awareness across emails/files, and model selection inside Studio for building more capable agents. (techcommunity.microsoft.com)Consumer Copilot’s new Smart mode
On the consumer side, Microsoft is introducing a new Copilot “Smart mode” that automatically adapts responses—optimizing for speed or deeper reasoning without manual model switching. Reporting indicates Smart mode runs on GPT‑5 and is available to free Copilot users on the web and apps. If you see “Smart” in the chat mode picker, you’re on the latest model. (theverge.com, windowscentral.com)GitHub Copilot
GitHub confirmed GPT‑5 is in public preview for all paid GitHub Copilot plans. It’s selectable via the model picker in GitHub Copilot Chat on github.com, Visual Studio Code (Ask/Edit/Agent), and GitHub Mobile. Org admins must enable the new GPT‑5 policy to expose it to users. (github.blog)Azure AI Foundry
Microsoft made the full GPT‑5 family generally available in Azure AI Foundry. Developers can use the Foundry model router to pick the “right” GPT‑5 tier automatically—Microsoft says this can cut inference cost by up to 60% while preserving fidelity—then scale into the Agent Service (with browser automation and MCP integrations) as it arrives. Data residency options span US/EU “Data Zones.” (azure.microsoft.com)GPT‑5 lands on Perplexity and Comet
Perplexity stated that GPT‑5 is now available on Perplexity and in the Comet AI browser for Max and Pro subscribers. Comet has been rolling out as an AI‑native, Chromium-based browser with early access for the $200/month Max plan. (thetechoutlook.com, windowscentral.com)Why Windows and Microsoft 365 users should care
- Better day‑to‑day answers in Windows 11’s Copilot and on the web: Smart mode reduces “which model do I pick?” friction, improving both quick web tasks and deep dives. (theverge.com)
- Stronger “work graph” understanding in Microsoft 365 Copilot: GPT‑5 can reason over your mail, docs, and calendars with improved context handling. (techcommunity.microsoft.com)
- Developer gains in GitHub Copilot and VS Code: clearer step‑by‑step plans, longer chains of tool calls, and improved code quality/outlining in longer sessions. (github.blog)
- Enterprise control via Azure AI Foundry: model routing, governance, telemetry, and residency help move pilots to production. (azure.microsoft.com)
Early performance and pricing details
OpenAI’s developer note touts state‑of‑the‑art results on real‑world coding benchmarks (e.g., 74.9% on SWE‑bench Verified), stronger tool‑calling, and more stable agentic behavior. For businesses, the headline is fewer factual errors on fact‑seeking prompts and tighter steerability. For API users, GPT‑5, 5‑mini, and 5‑nano allow trading depth for cost/latency. (openai.com)- API highlights: “minimal reasoning” knob, verbosity control, preamble messages before tool calls, and more robust custom tool schemas. (openai.com)
- ChatGPT availability: “Available to everyone” today; ChatGPT Team is live now; Enterprise/Edu are dated August 14. (openai.com)
How to try GPT‑5 today
- Microsoft 365 Copilot: The rollout is automatic; open Word/Outlook/Teams and look for improved answers and reasoning. Builders can select GPT‑5 in Copilot Studio. (techcommunity.microsoft.com)
- Copilot (consumer): Visit copilot.microsoft.com or the Copilot app and switch to “Smart” mode if offered; this routes your prompts to GPT‑5. (theverge.com)
- GitHub Copilot: In VS Code or github.com, open Copilot Chat and choose GPT‑5 in the model picker. Enterprise admins must enable the GPT‑5 policy. (github.blog)
- Azure AI Foundry: In Foundry Models, deploy a GPT‑5 variant and enable the model router to balance cost, latency, and quality. (azure.microsoft.com)
- Perplexity/Comet: If you’re on Max or Pro, GPT‑5 should appear automatically; no extra configuration required. (thetechoutlook.com)
Strengths worth calling out
- Unified reasoning + generation: GPT‑5 is designed to “think harder when needed,” minimizing mode‑switching and raising baseline answer quality—ideal for Smart mode in Copilot. (openai.com, theverge.com)
- Coding productivity: From SWE‑bench gains to clearer multi‑step plans, GitHub Copilot users should see fewer stalls and better end‑to‑end outcomes. (openai.com, github.blog)
- Enterprise readiness: Foundry’s router and agent roadmap, combined with data residency options, make GPT‑5 easier to govern at scale. (azure.microsoft.com)
Risks, open questions, and what to watch
- Opaque routing: Smart mode and Foundry’s router are powerful—but they’re automatic. Enterprises should require logging that shows which model handled which step for auditability and compliance. Microsoft’s posts tout savings and quality, but your governance team will want proof and override controls. (azure.microsoft.com, theverge.com)
- Cost dynamics: GPT‑5 output tokens aren’t cheap. Teams should tune verbosity and “reasoning effort” to rein in spend, and measure router claims against real workloads. (openai.com)
- Rollout variance: Availability differs by product and plan—e.g., ChatGPT Team today, Enterprise/Edu on August 14; GitHub Copilot requires an admin policy toggle; consumer Copilot’s Smart mode may stagger by region/app. (openai.com, github.blog, theverge.com)
- Platform lock‑in: Offloading crucial workflows to GPT‑5 deepens reliance on closed models and Azure. Mitigate with architecture that isolates prompts, tools, and data, enabling future model portability. (azure.microsoft.com)
Bottom line for Windows power users and IT
GPT‑5 meaningfully upgrades the Copilot experience across Windows and Microsoft 365 while giving developers and IT a cleaner path from prototype to production via Azure AI Foundry. If you live in GitHub Copilot or M365 all day, expect crisper planning, sturdier reasoning, and faster task completion—tempered by the need to monitor costs and maintain traceability as “smart” routing takes over more decisions under the hood. For now, turn it on, measure it, and instrument it—GPT‑5’s biggest win in the Windows ecosystem may be that you won’t have to think about picking the “right” model anymore. (techcommunity.microsoft.com, github.blog, azure.microsoft.com, theverge.com)Source: The Tech Outlook ChatGPT-5 launches across Microsoft 365 Copilot, Copilot, GitHub Copilot and Azure AI Foundry; Also available on Perplexity and Comet for Max and Pro users - The Tech Outlook
