what is mai code 2? price, benchmarks and where to use it
mai code 2 is microsoft ai's new coding model, announced on november 17, 2026 together with a lightweight mai code 2 flash. both are available in github copilot to eligible subscribers on annual plans, where github's docs list promotional prices for both, shown in the price table below.
microsoft publishes its own benchmark comparisons for mai code 2, covered below. this guide follows microsoft ai's announcement, github's changelog and github's pricing docs as we read them.
download the ReMynd desktop app to search text you saw in a focused window, by app and date.
what are mai code 2 and mai code 2 flash?
they are two coding models from microsoft ai. microsoft's announcement says they are built for everyday enterprise coding tasks.
| model | what microsoft says it is for |
|---|---|
| mai code 2 | taking code to production, deep debugging, multi-file implementation, repo-wide reasoning and ui |
| mai code 2 flash | high-volume work such as code explanations, targeted edits, routine fixes and rapid iterations |
microsoft adds a limit of its own: the models are "explicitly not yet tuned for harder problems", and it names fundamental science research and very long-horizon agentic work.
where can you use mai code 2?
github's changelog lists the plans and the places you can pick the models, and says both models are available to eligible github copilot subscribers on annual plans.
| model | github copilot plans |
|---|---|
| mai code 2 | pro, pro+, max, business and enterprise |
| mai code 2 flash | free, student, pro, pro+, max, business and enterprise |
to start using it, github's changelog points to these steps:
- find your copilot plan in the table above. github says both models are available to eligible github copilot subscribers on annual plans.
- on copilot business or enterprise, your plan administrator can manage access through the model policy in copilot settings, as described below.
- open the model picker in visual studio code, visual studio, copilot cli, the github copilot cloud agent, the github copilot app, github.com, github mobile for ios and android, jetbrains, xcode or eclipse, and select mai code 2 or mai code 2 flash.
for copilot business and enterprise, github says that under default model enablement, new models are enabled automatically unless an administrator has turned off the global default or explicitly disables the model.
microsoft's announcement says mai code 2 and mai code 2 flash are available in github copilot, that mai code 2 flash is in limited-access preview in microsoft foundry, and that microsoft will soon release them on openrouter, vercel, hermes, opencode and kilo.
how much does mai code 2 cost in github copilot?
github's models and pricing page lists prices per 1m tokens, and says 1 ai credit equals $0.01. the table below shows github's list prices for the mai models and four of the models microsoft compares them with.
| model | input | cached input | output |
|---|---|---|---|
| mai code 2, until may 17, 2027 | $0.50 | $0.025 | $2.50 |
| mai code 2, from may 17, 2027 | $1.00 | $0.05 | $5.00 |
| mai code 2 flash, until may 17, 2027 | $0.05 | $0.005 | $0.25 |
| mai code 2 flash, from may 17, 2027 | $0.10 | $0.01 | $0.50 |
| gpt-6.1 sol, up to 272k tokens | $2.00 | $0.10 | $10.00 |
| claude opus 5.5 | $4.00 | $0.20 | $20.00 |
| gpt-6 luna, up to 272k tokens | $0.10 | $0.01 | $0.50 |
| claude haiku 5.5, up to 100k tokens | $0.10 | $0.01 | $0.50 |
github's docs add that copilot pro and pro+ subscribers on existing annual billing plans using the request-based billing model have different model multipliers.
what does microsoft claim about performance?
microsoft's announcement gives these results; the first row is averaged over four benchmarks that target everyday enterprise coding tasks in github copilot (deepswe 1.1, terminal bench 2.1, swe-atlas and frontiercode v1.1).
| comparison | what microsoft says |
|---|---|
| mai code 2 vs gpt-6.1 sol | 98% of the performance at 40% of the cost |
| mai code 2 vs claude opus 5.5 | 96% of the performance at 5% of the cost |
| mai code 2 flash vs gpt-6 luna and claude haiku 5.5 | comparable performance at 65% and 20% of the cost, across three benchmarks |
| mai code 2 vs gpt-6 sol, blind frontend app evaluations | developers preferred mai code 2's results 64% of the time |
microsoft's notes say all models were run at max reasoning effort, three of the benchmarks ran on its internal eval platform msbench with the github copilot cli without subagents, frontiercode scores and costs are from artificial analysis, and the costs of mai code 2 and mai code 2 flash are shown at their 50% promotional prices. the visual coding result comes from a total of 258 annotators.
why should you treat these numbers carefully?
microsoft's own notes say results reflect evaluations on specific coding tasks and configurations, and that performance and cost comparisons vary by workload, model configuration, reasoning effort and pricing assumptions.
these are microsoft's numbers, not an independent test. we did not run mai code 2 on any benchmark, and microsoft's cost ratios use the promotional prices of the two mai models, which github's docs say end on the date in the price table.
where does ReMynd fit?
ReMynd records your screen on a mac, runs ocr locally with apple vision, and keeps a searchable index on your computer. text comes from the focused window only, and video covers windows you have not excluded.
in our view, a coding agent fails less often because it cannot write code, and more often because it does not know what you were doing, and a cheaper model does not fix that. ReMynd's agent window runs your own claude code or codex cli against that history, with read-only access, and connected agents send what they read to their own cloud. on a mac it needs macos 14 or later on an apple silicon mac.
ReMynd also runs on windows 11 (x64 or arm64).
for more on model choice, see what is the best ai model for coding and ai model pricing compared.
limits of this guide
- it follows microsoft ai's announcement, github's changelog and github's pricing docs as read on november 19, 2026. we did not test mai code 2.
- github's prices are list prices for github copilot's usage-based billing. other places microsoft names, such as microsoft foundry, were not checked for prices.
- the benchmark results are microsoft's own. we did not verify deepswe, terminal bench, swe-atlas or frontiercode scores independently.
- model names, plans and prices change often, so check the pages for the current ones.
download the ReMynd desktop app and search what you saw in a focused window, by app and date.