Skip to main content

Current model · Microsoft AI

MAI Code 2 Flash benchmarks, pricing and specs

MAI Code 2 Flash is a Microsoft AI model released on November 17, 2026 that takes text and image input and produces text. It costs $0.05 per 1M input tokens and $0.25 per 1M output tokens and accepts up to 1,050,000 tokens of context.

Per 1M input tokens
$0.05Listed today
Per 1M output tokens
$0.25Listed today
Context window
1.05M#1 of 75 current models

Data verified November 18, 2026 · Dataset 2026-11-18

Share this

Specs

Key facts

Provider
Microsoft AI
Model family
MAI Code
Release date
November 17, 2026
Status
Current
Access
limited access
API model id
Not reported
Context window
1,050,000 tokens
Max output
128,000 tokens
Input
text, image
Output
text
Task families
text, multimodal
Declared capabilities
reasoning, coding, agentic coding, image understanding

Charts

Against every current text model

Where MAI Code 2 Flash lands among the current models of its task, on every independent benchmark, on price and on context.

Where it stands

0 of 64 list a lower price per 1M input tokens. Each bar is the share of current models it beats.

MAI Code 2 Flash among the current text models, one row per benchmark, price and context window.

Price and context

Price per 1M input tokens

#1of 64

$0.05 · cheaper than 100%

Price per 1M output tokens

#1of 64

$0.25 · cheaper than 100%

Context window

#1of 75

1.05M tokens · longer than 81%

Price

MAI Code 2 Flash pricing

It is the cheapest of 64 current text models priced per 1M input tokens.

UnitPrice (USD)TierAppliesSource
1M input tokens$0.05standardUntil May 16, 2027
1M input tokens$0.10standardFrom May 17, 2027
1M output tokens$0.25standardUntil May 16, 2027
1M output tokens$0.50standardFrom May 17, 2027
1M input tokens$0.005cachedUntil May 16, 2027
1M input tokens$0.01cachedFrom May 17, 2027

Benchmarks

Independent benchmark results

Scores come from independent evaluators that ran the model themselves, pinned to a dated snapshot. Rank is among current models in the same snapshot. Microsoft AI's own launch figures are in the next section, because they are not measured the same way.

No independent benchmark has published a result for MAI Code 2 Flash yet. It is still compared below on price, context and capabilities.

Provider results

Reported by Microsoft AI

Microsoft AI's own results for MAI Code 2 Flash, beside the rivals it chose, as its table lists them. Labs pick the benchmarks and run them their own way, so compare within one table only. Internal benchmarks and other evaluators' indexes are left out.

MAI Code 2 and MAI Code 2 Flash launch charts, November 17, 2026Microsoft AI · 2026-11-18

BenchmarkMAI Code 2MAI Code 2 FlashGPT-6.1 SolClaude Opus 5.5GPT-6 LunaClaude Haiku 5.5
DeepSWE 1.173.2%59.9%74.6%69.0%59.6%46.0%
Terminal-Bench 2.184.6%78.3%85.0%78.7%69.3%73.4%
SWE Atlas43.1%44.8%47.4%57.5%36.4%42.1%

Microsoft's two charts from the MAI Code 2 announcement: MAI Code 2 against GPT-6.1 Sol and Claude Opus 5.5, and MAI Code 2 Flash against GPT-6 Luna and Claude Haiku 5.5; the model card gives the same scores for the two MAI models. Microsoft ran every model at max reasoning effort on MSBench, its evaluation platform, in the GitHub Copilot CLI harness without subagents. Left out here: FrontierCode v1.1, whose scores Microsoft takes from Artificial Analysis, the charts' averages, and the blind front-end preference test.

Fit

Strengths, cautions and best uses

Strengths
  • The small, fast MAI Code 2 model for explanations, targeted edits and routine fixes: 12B active and about 398B total parameters, with the same 1,050,000-token context and 128,000-token output as MAI Code 2
  • In GitHub Copilot it costs $0.05 input, $0.005 cached input and $0.25 output per 1M tokens until May 17, 2027, half of GPT-6 Luna's and Claude Haiku 5.5's list rates
  • In Microsoft's runs (GitHub Copilot CLI, max effort) it scores 78.3% on Terminal-Bench 2.1, 59.9% on DeepSWE 1.1 and 44.8% on SWE Atlas, ahead of GPT-6 Luna (69.3%, 59.6% and 36.4%) and Claude Haiku 5.5 (73.4%, 46.0% and 42.1%)
Cautions
  • Limited access: in GitHub Copilot on every plan, Free and Student included, and in a limited-access preview in Microsoft Foundry, with no public API yet
  • The Copilot rates double on May 17, 2027, to $0.10 input and $0.50 output per 1M tokens, the same as GPT-6 Luna and Claude Haiku 5.5
  • Its DeepSWE lead over GPT-6 Luna is 0.3 points in Microsoft's own runs, and no independent benchmark has scored it yet
  • Replaces MAI-Code-1.1-Flash, which GitHub retires from Copilot on December 18, 2026
Best for
  • High-volume coding help in Copilot: explanations, targeted edits, routine fixes and fast iteration
  • Teams on Copilot Free or Student plans that want a current coding model

From benchmark to production

Building on MAI Code 2 Flash?

Public scores rarely hold on your own data. We test it against the alternatives on your real tasks, then build the agent or pipeline around the model that wins.

AI Model Cost AuditLaunch price $100, regular price $150

Alternatives

Alternatives to MAI Code 2 Flash

The closest current text models by price. The cheapest model that scores as well or better is listed first; failing that, the cheapest one within 3% of the benchmark's scale.

On score and price

MAI Code 2 Flash beside the 4 alternatives listed below, on Arena Text and on blended price per 1M tokens.

MAI Code 2 Flash and its alternatives on score and price.

Arena Text

Measured November 11, 2026, higher is better

MAI Code 2 Flash
Not scored
Hy4 flash
Not scored
Claude Haiku 5.5
1,421
GPT-4.1 nano
1,322
GPT-6 Luna
1,441

Blended price per 1M tokens

Listed today, lower is better

MAI Code 2 Flash
$0.1
Hy4 flash
$0.123
Claude Haiku 5.5
$0.2
GPT-4.1 nano
$0.175
GPT-6 Luna
$0.2

Further reading

More on MAI Code 2 Flash

Other MAI Code models

FAQ

MAI Code 2 Flash FAQ

How much does MAI Code 2 Flash cost?

MAI Code 2 Flash costs $0.05 per 1M input tokens and $0.25 per 1M output tokens at the standard tier. It is the cheapest of 64 current text models priced per 1M input tokens. Source: GitHub, verified November 18, 2026.

What is the context window of MAI Code 2 Flash?

MAI Code 2 Flash accepts up to 1,050,000 tokens of context. It can return up to 128,000 tokens in one response. That ranks #1 of 75 current text models by context length.

What benchmark results has Microsoft AI published for MAI Code 2 Flash?

In the MAI Code 2 and MAI Code 2 Flash launch charts (November 17, 2026), Microsoft AI reports 59.9% on DeepSWE 1.1, 78.3% on Terminal-Bench 2.1 and 44.8% on SWE Atlas. These are Microsoft AI's own figures, beside the rivals it chose, not independent measurements.

Is MAI Code 2 Flash open source or API only?

Neither. MAI Code 2 Flash has no public API and no published weights; Microsoft AI offers it only through the limited access described in this page's cautions.

What are the best alternatives to MAI Code 2 Flash?

The closest current alternatives by price and measured performance are Hy4 flash (40% more per 1M input tokens, $0.07 vs $0.05), Claude Haiku 5.5 (2.0x the price per 1M input tokens, $0.10 vs $0.05), GPT-4.1 nano (2.0x the price per 1M input tokens, $0.10 vs $0.05) and GPT-6 Luna (2.0x the price per 1M input tokens, $0.10 vs $0.05).

AI Model Brief

The AI models worth knowing, once a week

What shipped, how it scores on independent benchmarks, what it costs per million tokens and what was deprecated, from the data behind our AI model benchmarks.

One email a week. We send a confirmation link first, and every issue has a one-click unsubscribe.

We use your address only to send the brief. See our privacy policy.

See this week’s brief

Sources

Sources

All 300 sources behind the benchmarks, with their versions, licenses and notes, are in the source catalog.

Cite this page

Van Data Team. "MAI Code 2 Flash Benchmarks, Pricing and Specs." AI Model Benchmark Dataset 2026-11-18. https://vandatateam.com/ai-model-benchmarks/mai-code-2-flash

Reuse the data: JSON · CSV · or explore every model in the AI model benchmark comparison.

Production evaluation

Deciding whether MAI Code 2 Flash fits your workload?

Public benchmarks narrow the shortlist. We test the finalists on your own tasks, data and cost limits, then build the routing and review controls to run the winner in production.

  • Evaluation cases and acceptance thresholds tied to real work
  • Model routing, fallbacks and cost controls