PlatformsThe briefing
xAI says Grok 4.7 is its most capable model for coding and knowledge work and kept the entry price of Grok 4.6. Access is split across the API, Cursor, Grok Build and a gradual GitHub Copilot rollout.
On September 21, 2026, xAI launched Grok 4.7. xAI calls it its most capable model for coding and knowledge work. It says the model works longer on hard problems and checks its own work more carefully than Grok 4.6 did.
The price is the same as Grok 4.6 at the entry level. xAI's documentation lists two dollars per million input tokens and six dollars per million output tokens for shorter prompts. A token is a small chunk of text. Once a prompt passes 200,000 tokens, both rates double. Long agent sessions can cross that line, so the headline price may not be the price you pay.
There are two versions, and you cannot use both everywhere. The standard model, grok-4.7, is on the xAI API and on its US regional endpoint. A fast version, grok-4.7-fast, costs twice as much. xAI says it is available only in Cursor and Grok Build, not on the public API.
GitHub says Grok 4.7 is rolling out in Copilot on the Pro, Pro+, Max, Business and Enterprise plans, in editors such as Visual Studio Code, JetBrains and Xcode. GitHub says the rollout is gradual. For Business and Enterprise, new models switch on automatically unless an administrator has turned off the default or blocked this model.
Compare like with like. A result from the fast version inside Cursor tells you little about the standard model on the API, and the two cost different amounts. xAI also documents a reasoning setting with four levels, with high as the default. That setting changes cost, so log it for each run. Then measure one real agent session. Count the prompt size at each step and see how often it passes 200,000 tokens before you estimate a bill.
The main decision is about control. On Copilot Business and Enterprise, a new model can reach your developers without anyone choosing it. Ask your administrator whether the default for new models is on or off. Also note that xAI names US regional endpoints. It does not say what covers other regions, and a missing region means unknown. If code or customer data would go to a new provider, run it through the review you already use for vendors before habits form.
Ask your Copilot administrator for the current model policy. Write down whether new models switch on by default, which plan your seats use and which editors are affected. Then count prompt tokens per step in one real agent session and compare them with the 200,000 limit. Both checks can be undone, and neither commits you to the model.
Source material behind this briefing. Our practical implications are Ashlr’s interpretation.
This briefing was prepared with AI assistance from the linked sources. The analysis may include interpretation; consult the original sources for complete context. Cover artwork is illustrative.
Andy Jassy told analysts that frontier model competition will produce several comparably capable systems. Bedrock already routes to more than 100 models through one API, and most traffic now runs on Trainium.
PlatformsThe briefing
AI-generated illustration
Mistral AI has made the weights of its 1-trillion-parameter multimodal model available under open release, enabling sovereign control and local deployment options.
PlatformsThe briefing
AI-generated illustration
AWS released general availability of training sessions in SageMaker AI serverless model customization. Builders can now write their own post-training loops, including custom loss functions and multi-turn RL, while SageMaker handles GPU allocation, distributed execution, and fault recovery.
