GitHub Copilot continues evolving from a code completion tool into a system where users also need to manage models, costs and agents.
GitHub added three profiles for automatic model selection:
- Efficiency
- Balance
- Intelligence
The goal is to let users choose how much weight Copilot should give to cost, speed and quality.
What changed in model selection
Instead of users always choosing a specific model, Copilot can choose one automatically.
The three modes use the same pool of available models but weigh these differently:
- Price
- Speed
- Quality
Efficiency is intended to put more weight on efficiency.
Intelligence is intended to put more weight on quality.
Balance tries to combine them.
What this changes for developers
AI coding tools are becoming systems combining several models.
Users no longer always care exactly which model is behind each action.
But they do care:
- How long it takes
- How many AI credits it consumes
- And whether the result is good enough.
The problem is that when the system chooses automatically, it becomes harder to know in advance what cost and behavior to expect.
Code Review is changing too
GitHub also updated Copilot Code Review.
When reviewing a Pull Request again, Copilot can now mark comments already addressed as resolved and leave unresolved problems open.
Copilot can also suggest a commit message when users implement its recommendation.
In addition, the review system can use shell commands to examine changes.
Lite reviews can combine findings from multiple agents.
Code Review as an ongoing agent
This is a small interface change but a large conceptual one.
Traditional Code Review is a list of comments.
An agent-based system can:
- Remember what has already been fixed
- Leave problems open
- Run tools
- Check the result
- And combine information from several processes.
In other words, review turns from a one-time response into an ongoing process.
How much GitHub Copilot costs today
According to GitHub's official documentation:
- Copilot Free: free
- Copilot Pro: 10 dollars per month
- Copilot Pro+: 39 dollars per month
- Copilot Max: 100 dollars per month
- Copilot Business: 19 dollars per user per month
- Copilot Enterprise: 39 dollars per user per month
In the new plans, some usage is measured through AI credits.
GitHub states:
- Pro: includes 1,500 AI credits per month
- Pro+: includes 7,000
- Max: includes 20,000
- Business: includes 1,900 per user in an organizational pool
- Enterprise: includes 3,900 per user
Additional consumption in Business and Enterprise plans is priced at 0.01 dollars per AI credit.
Has the price changed now?
No.
At publication, no change to list prices was found.
The news is mainly about how Copilot selects models and consumes resources.
What about Israel?
GitHub's documentation has no separate availability statement for Israel.
Conversation quality in Hebrew is not explicitly documented either.
Distinguish GitHub's support for programming languages from support for a natural language such as Hebrew.
What GitHub has not yet published
GitHub does not publish the routing mechanism in full.
It is not clear in advance which model will be selected for each action.
It is unknown exactly how much each of the three modes will save or cost.
There are also no independent tests yet showing whether combining multiple agents in Code Review actually detects more bugs.