Service eligibility and regional restrictions
PuppyIP serves only compliant overseas businesses and their authorized personnel. Proxy services are not available in mainland China. The service may only be used for lawful business activities outside mainland China. Use of this service within mainland China is prohibited.
Hosting a proxy IP or server overseas does not change these restrictions. The service must not be provided to end users in mainland China through relaying, forwarding, sharing or resale. Before use, read the Terms of Service.
Key Takeaways
- The November 17 announcement lists Code 2 for Pro, Pro+, Max, Business and Enterprise. Flash also includes Free and Student, and eligible annual subscribers can access both models.
- Current general access documentation still says Free and Student have Auto model selection only. Flash plan eligibility does not establish a manual picker for every account.
- Both models use token-based billing. Promotional rates change on May 17, 2027, with separate input, cached-input and output prices; one AI credit equals $0.01 USD.
- Business and Enterprise administrators still control model access. Check client support, account options, budgets and actual usage separately.
When should you choose Code 2 or Flash?
GitHub announced the two Microsoft AI coding models in Copilot on November 17, 2026. The current supported-model table marks both as GA, meaning generally available. MAI Code 2 targets deep debugging, implementation across files, refactoring and reasoning across a repository. MAI Code 2 Flash targets code explanations, focused edits, routine fixes and rapid iteration.
These are product descriptions, not guarantees that every project will be faster or cheaper. For example, you might give Flash a clearly located null-handling fix, then consider Code 2 for tracing calls across files. This is a hypothetical selection example, not a comparison run for this guide. GitHub also describes MAI models as continuously improving, so performance and behavior may change with new checkpoints.
Which plans qualify, including annual subscriptions?
The announcement lists MAI Code 2 for Copilot Pro, Pro+, Max, Business and Enterprise. MAI Code 2 Flash also includes Copilot Free and Student. It explicitly includes eligible annual subscribers for both models, so restrictions described in older guides for other models should not simply be carried over. Check the account entitlement and billing status to establish whether a particular older plan qualifies.
An access limitation belongs alongside that eligibility: current general GitHub model-access documentation still says Free and Student have Auto model selection only. Flash inclusion on those plans does not establish a separate manual Flash option for every free or student account. Inspect the Auto or model options actually available to that account. These sources give no date for completing access for every account.
Where can you find the models, and which versions are required?
GitHub lists Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains, Xcode and Eclipse. Accounts that support manual selection can check the corresponding model picker; Free and Student users should retain the Auto limitation described above.
At this review, GitHub lists Visual Studio 17.14.6 as the minimum IDE version for both models. The corresponding VS Code, JetBrains, Xcode and Eclipse entries remain TBD, meaning not yet specified. Do not invent a shared minimum version or identical menus across clients. Record the client, version and signed-in account, then compare them with the official support table.
Which model policy should an administrator check?
Copilot Business and Enterprise administrators can control MAI access through the model policy in Copilot settings. Under the announcement’s default model-enablement rule, new models are enabled automatically unless an administrator has turned off the global default or explicitly disabled the model. Have an authorized administrator check the applicable enterprise, organization and individual model settings.
Default model availability controls permission to use a model. It is not a spending cap and does not establish that existing tasks switched models. Check the model actually selected for the current session, then decide whether to use it under the organization’s data and budget requirements. The default for new models and the separate default policy for new features are different controls.
When do promotional rates end, and what follows?
GitHub’s current table prices each one million tokens in USD. Tokens measure the input and output a model processes; cached input is context reused by the model. During the current promotion, MAI Code 2 costs $0.50 for input, $0.025 for cached input and $2.50 for output per million tokens. Flash costs $0.05, $0.005 and $0.25 respectively.
The pricing footnotes say that from May 17, 2027, Code 2 rates become $1.00, $0.05 and $5.00; Flash rates become $0.10, $0.01 and $0.50 in the same order. They do not specify the exact changeover time or time zone. Recheck official rates and billing near that date. These are Copilot prices reviewed on November 18, not a quote for a separate API or local execution.
How do token rates translate into a usage estimate?
One GitHub AI credit equals $0.01 USD. Included allowances vary across individual plans, while Business and Enterprise pool per-seat allowances at the billing entity level. Consumption depends on the selected model and token counts. Access through a plan does not mean unlimited use, and a token rate is not a fixed task price.
Suppose one interaction uses exactly 100,000 input tokens and 20,000 output tokens, with no cached input, and you count only these two model-token components. At promotional rates, Code 2 costs $0.05 + $0.05 = $0.10, or 10 AI credits. Flash costs $0.005 + $0.005 = $0.01, or 1 AI credit. This is a stated arithmetic example, excluding additional calls, tool steps or other charges; it does not predict a real task’s complete bill.
Before starting, check the personal or organizational allowance, additional usage and budget-stop settings. Review actual usage afterward. Whether lower token rates reduce the total also depends on context, rework and the number of calls. This guide measured no savings percentage.
What should you check if a model is missing?
First check the signed-in account, actual Copilot plan and annual-subscription eligibility, followed by the client and any published version requirement. For a managed account, check administrator model policies next. Free and Student users should inspect Auto rather than treating a missing manual Flash menu as a fault. Record the options actually shown; an announcement is not proof of access for a particular account.
These sources do not provide a complete country-by-country availability table or establish identical access for every account. If the conditions and interface still disagree, keep the client version, plan and non-sensitive error details, then use GitHub’s support process. A proxy exit cannot add subscription eligibility, change organization policies or bypass billing. This guide did not sign in, enable models or make paid calls.
Sources
Frequently Asked Questions
Does this mean MAI Code 2 can be downloaded for offline local use?
That is not established here. Current access documentation separately says eligible Windows users can add local MAI Code 1.1 Flash through supported entry points when their hardware meets the requirements. That is a different model name. This Copilot announcement and its token rates do not establish local downloads or offline execution for Code 2 or Code 2 Flash.
Do both models support a one-million-token context window and reasoning levels?
GitHub’s current support table lists MAI Code 2 and Flash as supporting a one-million-token context window and configurable reasoning levels. The documentation limits that context window to VS Code and Copilot CLI. Larger context windows or higher reasoning levels consume more tokens and AI credits. Model support does not establish identical options in every client.