Skip to content

OpenAI

Alphabet Q3, OpenAI API credits and Copilot Fusion

3 min read

Written by an AI model from the linked sources and checked against them by a second model. How we work

All about: OpenAI models and products →
Image: pexels.com

GitHub launched Copilot Fusion and pitches it against Claude Code by name, saying it nearly matches Claude Code at about half the cost, according to an October 29 AI roundup on Threads. If you pay for ChatGPT Pro or Business, there is also an item you can act on now: OpenAI lets those users claim up to $500 a month in API credits.

How Fusion picks models

Fusion replaces a specific model in the picker and decides, prompt by prompt, which models to use and how they work together, says a DEV Community breakdown that dates general availability to October 28. You only set the effort: Standard, High or Max. Fusion then either hands the task to one model, or lets an efficient model draft and escalates to a stronger one when a quality check rejects the draft, or has a second model review a draft before the first one revises it. It only uses models your plan and your organisation’s policies allow, and GitHub does not publish which ones it picks for a given task.

The half-cost claim comes from the Artificial Analysis Coding Agent Index. There, Fusion Max scored 60, “within two points” of Claude Code running Opus 5.5 Max, at about half the average API cost per task, with about half as many tokens and about 40% less time per task, the post reports. So the result belongs to the top tier. Standard comes with every Copilot plan, while High and Max need Pro+, Max, Business or Enterprise.

What you pay for it

Fusion has no fee of its own. Each model it calls is billed at that model’s standard rate out of GitHub AI Credits, and through November 30 paid Copilot plans get 30% off model costs when they use Fusion, per the DEV Community write-up. Those credits replaced premium request units on June 1, when GitHub moved Copilot to usage-based billing, Morph notes in its comparison.

The benchmark measures cost as API spend per task. That maps directly onto a Copilot bill, but not onto a Claude subscription: Claude Code comes with every paid Claude plan and shares one usage pool with Claude chat, and pay-as-you-go API billing is the other way to pay for it.

Projects, Automations and the earnings

Fusion was not GitHub’s only agent release. Copilot Projects orchestrates multiple agents across repos, and Copilot Automations runs jobs on demand, on a schedule or on GitHub activity. GitHub also added a Copilot Security extension and a feature called Copilot dreaming.

Mistral’s Responses API for building agents is now in public preview, and the Grok API accepts guidelines that steer how it handles content categories. Nvidia released Nemotron 3.5 Transcribe, an open-weight transcription model covering 40 languages. OpenAI’s API docs reference an unannounced GPT Image 2.6 with markup-guided editing, and in the US the OpenAI API, Codex and ChatGPT Work are now available on Oracle Marketplace.

Alphabet’s Q3 revenue rose 25% to $127.6 billion, with Cloud up 103% and Gemini at 1 billion users. Microsoft reported $92.3 billion for its fiscal first quarter, with Azure up 48%. In that same quarter, which ended September 30, a regulatory filing shows Microsoft booked $10 billion in revenue from its OpenAI arrangements, revenue-sharing payments included.

No comments yet. Be the first.

Join the conversation

Sign in to leave a comment. It appears after moderation.