Microsoft · Released, not on our provider
MAI Code 2 Flash (2026): what Microsoft’s lightweight coding model is, its price, who gets it and what to use instead
MAI Code 2 Flash is Microsoft AI’s lightweight, low-cost coding model, announced on November 17, 2026 with MAI Code 2 and priced in GitHub Copilot at a promotional $0.05 per million input tokens until May 17, 2027, but it is not on OpenRouter, so this site cannot run it.
Use it now
Chat
This chat starts with Claude Haiku 5.5, not MAI Code 2 Flash.
≈ 1 credit per reply · See all models
New accounts get 5 free credits once, valid for 7 days. They work on the lower-priced models here; the priciest ones need credits you bought.
Start chatting — 5 free credits- Maker
- Microsoft
- Type
- Text
- Released
- 2026-11-17
- OpenRouter id
- None
Checked
MAI Code 2 Flash: facts, with sources
- Microsoft AI announced MAI Code 2 Flash together with MAI Code 2 on November 17, 2026, in a post titled Everyday coding at scalable prices, and says both are available in GitHub Copilot. It describes Flash as the model for high-volume work such as code explanations, targeted edits, routine fixes and rapid iterations, at a much lower cost.Source: microsoft.ai · checked 2026-11-19
- GitHub’s models and pricing table lists MAI Code 2 Flash as generally available in the Lightweight category at $0.05 per million input tokens, $0.005 per million cached input tokens and $0.25 per million output tokens, a promotional price until May 17, 2027; from May 17, 2027 it is $0.10, $0.01 and $0.50. GitHub turns token use into AI credits (1 AI credit is $0.01): plans include credit allowances, and use beyond them is billed at these per-token rates.Source: docs.github.com · checked 2026-11-19
- GitHub’s changelog lists MAI Code 2 Flash for Copilot Free, Student, Pro, Pro+, Max, Business and Enterprise, "to eligible GitHub Copilot subscribers on annual plans", billed under usage-based billing. You pick it in the model picker of VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the Copilot app, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse; Business and Enterprise administrators control access through the model policy in Copilot settings.Source: github.blog · checked 2026-11-19
- Microsoft’s announcement says MAI Code 2 Flash is in limited-access preview in Microsoft Foundry, and that it will release the new models "soon" on OpenRouter, Vercel, Hermes, OpenCode and Kilo; it gives no date and does not say which plans or regions get Foundry access.Source: microsoft.ai · checked 2026-11-19
- Microsoft’s own claim, not tested by us: the announcement says MAI Code 2 Flash offers comparable performance to GPT-6 Luna and Claude Haiku 5.5 at 65% and 20% of the cost. Its footnote says all models ran at max reasoning effort, the DeepSWE 1.1, Terminal Bench 2.1 and SWE Atlas runs used MSBench with the GitHub Copilot CLI without subagents, and Flash’s cost is shown at its 50% promotional price. The announcement gives no Flash scores.Source: microsoft.ai · checked 2026-11-19
- Microsoft’s model page for MAI Code 2 says Flash is "designed for routine, high-volume, affordable work at scale", and its headline figure for Flash is 1/5 of the cost of Claude Haiku 5.5 and 12% higher performance (Microsoft’s own figure, not tested by us; the page names no benchmark for it).Source: microsoft.ai · checked 2026-11-19
- The announcement says both new models are "explicitly not yet tuned for harder problems, like fundamental science research and very long-horizon agentic work".Source: microsoft.ai · checked 2026-11-19
- The pages we read state no context window, maximum output or API model id for MAI Code 2 Flash, and the announcement does not say whether Flash accepts images (its visual-coding claim is about MAI Code 2). Microsoft’s separate page for Flash, microsoft.ai/models/mai-code-2-flash, returned "not found" when we read it, and we could not read the model card, so none of these is on this page.Source: microsoft.ai · checked 2026-11-19
- Before it: Microsoft’s announcement of MAI-Code-1-Flash is dated June 2, 2026, and says it was rolling out to GitHub Copilot individual users in Visual Studio Code. GitHub’s changelog says that with the launch of MAI Code 2 Flash, it will deprecate MAI-Code-1.1-Flash across all Copilot experiences on December 18, 2026, and suggests the latest MAI Code Flash model instead.Source: github.blog · checked 2026-11-19
- Microsoft’s MAI-Code-1-Flash announcement (June 2, 2026, updated June 8) calls it a coding model built for fast, efficient help in everyday developer workflows, and says it was trained from the ground up without distillation from third-party models.Source: microsoft.ai · checked 2026-11-19
- MAI Code 2 Flash is not on OpenRouter’s model list (checked November 19, 2026, 479 models): the list has no Microsoft MAI id, and the only Microsoft-prefixed id is microsoft/phi-4. This site cannot run MAI Code 2 Flash; models such as Claude Haiku 5.5, MiniMax M3.1 Flash and Hy4 Flash can be chatted with here.Source: openrouter.ai · checked 2026-11-19 · second-hand
Our table
MAI Code 2 Flash vs MAI Code 2: price, who gets it and what Microsoft claims
- GitHub describes MAI Code 2 as a powerful, general-purpose coding model for complex tasks such as deep debugging, multi-file implementation, refactoring and repository-wide reasoning, and MAI Code 2 Flash as a lightweight, cost-efficient model for code explanations, targeted edits, routine fixes and rapid iteration; both are meant for agentic coding workflows.Source: github.blog · checked 2026-11-19
- Microsoft’s own claims for MAI Code 2, not tested by us: averaged over four benchmarks (DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1), it gives 98% of the performance of GPT-6.1 Sol at 40% of the cost, and 96% of Claude Opus 5.5’s performance at 5% of the cost, with costs shown at its 50% promotional price.Source: microsoft.ai · checked 2026-11-19
Same launch, two models
MAI Code 2 Flash and MAI Code 2 side by side
MAI Code 2 Flash and MAI Code 2 launched on the same day and are priced in GitHub Copilot as the cheap and the strong option of one family. Flash costs a tenth of MAI Code 2 per token and reaches more plans, including Copilot Free and Student, but Microsoft publishes no Flash scores, only cost ratios. MAI Code 2 has no page of its own here; MAI Code 2 vs Claude Opus 5.5 covers it. For a cheap, fast model you can actually run on this site, see Claude Haiku 5.5.
| What | MAI Code 2 Flash | MAI Code 2 |
|---|---|---|
| GitHub category and status | Lightweight, GA | Powerful, GA |
| Input / cached input / output until May 17, 2027 (promotional) | $0.05 / $0.005 / $0.25 | $0.50 / $0.025 / $2.50 |
| Input / cached input / output from May 17, 2027 | $0.10 / $0.01 / $0.50 | $1.00 / $0.05 / $5.00 |
| GitHub Copilot plans | Free, Student, Pro, Pro+, Max, Business, Enterprise | Pro, Pro+, Max, Business, Enterprise |
| Microsoft Foundry | Limited-access preview | Not mentioned in the announcement |
| What GitHub says it is for | Code explanations, targeted edits, routine fixes, rapid iteration | Deep debugging, multi-file implementation, refactoring, repository-wide reasoning |
| Microsoft’s own cost and performance claim | Comparable performance to GPT-6 Luna and Claude Haiku 5.5 at 65% and 20% of the cost; no scores given | 98% of GPT-6.1 Sol’s performance at 40% of the cost; 96% of Claude Opus 5.5’s at 5% of the cost |
Source: docs.github.comSource: github.blogSource: microsoft.ai
Our table
MAI Code 2 Flash price next to MAI-Code-1.1-Flash and Claude Haiku 5.5 on GitHub
- On the same GitHub table MAI-Code-1.1-Flash is $0.20 / $0.02 / $1.20 (input / cached input / output, per million tokens), and Claude Haiku 5.5 is $0.10 / $0.01 / $0.50 for prompts up to 100K input tokens (plus $0.125 for cache writes), rising to $0.50 / $0.05 / $2.50 above 100K; all four are marked generally available and Lightweight. MAI Code 2 Flash’s promotional input and output prices are half of Haiku 5.5’s on this table, not a fifth, and the pages we read do not say how Microsoft’s 1/5 cost figure is worked out.Source: docs.github.com · checked 2026-11-19
GitHub price list
What MAI Code 2 Flash costs in GitHub Copilot against its predecessor and Claude Haiku 5.5
Microsoft says Flash costs 1/5 of Claude Haiku 5.5, but the per-token prices on GitHub’s table put its promotional price at half of Haiku 5.5 and its price from May 17, 2027 level with Haiku 5.5. The gap may come from how many tokens each model spends per task, which the pages we read do not show, so treat 1/5 as Microsoft’s figure, not ours. Other cheap models you can chat with here include MiniMax M3.1 Flash and Hy4 Flash.
| Per million tokens | MAI Code 2 Flash until May 17, 2027 | MAI Code 2 Flash from May 17, 2027 | MAI-Code-1.1-Flash | Claude Haiku 5.5 |
|---|---|---|---|---|
| Input | $0.05 | $0.10 | $0.20 | $0.10 |
| Cached input | $0.005 | $0.01 | $0.02 | $0.01 (cache write $0.125) |
| Output | $0.25 | $0.50 | $1.20 | $0.50 |
| GitHub category and status | Lightweight, GA | Lightweight, GA (same model) | Lightweight, GA; deprecated in Copilot on Dec 18, 2026 | Lightweight, GA |
| Output price as a share of Haiku 5.5’s (our arithmetic) | 50% | 100% | 240% | 100% (reference) |
Get told when it is available
Sign in and press the button: we store your account and this model, nothing else. No emails are sent yet.
Nearest available
Alternatives to MAI Code 2 Flash you can use now
Claude Haiku 5.5
Anthropic$0.10 / $0.50 (per million tokens)
- Anthropic’s model page says Claude Haiku 5.5 was released on October 7, 2026, is the latest Haiku, and will not be retired sooner than October 7, 2027. Anthropic’s launch post is dated the same day.Source: platform.claude.com · checked 2026-11-17
- In its launch post Anthropic calls Haiku 5.5 "the cheapest, fastest, and most capable small model we’ve ever released", for high-volume, cost-sensitive tasks such as summaries, compactions, database queries and classification, and as a coding subagent alongside Opus 5.5 and Sonnet 5.5. It calls Haiku 5.5 its first Haiku-class model with an adjustable effort setting.Source: anthropic.com · checked 2026-11-17
- Anthropic’s model page says it is built for high-volume, latency-sensitive work such as classification, routing, extraction and subagent tasks, and rates its latency "Fastest" in the current lineup (Fable 5.2 and Opus 5.5 are "Moderate", Sonnet 5.5 is "Fast").Source: platform.claude.com · checked 2026-11-17
MiniMax M3.1 Flash
MiniMax$0.15 / $0.60 (per million tokens)
- MiniMax’s launch post, dated October 19, 2026, says MiniMax is officially rolling out M3.1 Flash that day, and MiniMax’s API release notes have an entry for it on the same date. The post says an early version ran anonymously as "Space Bunny" before the launch.Source: minimax.io · checked 2026-11-17
- MiniMax calls it a "next-generation Flash model" and its API release notes describe it as an open-weight model for everyday work. MiniMax’s own claims, not tested by us: overall cost for agentic tasks about 70% lower than its (unnamed) predecessor, online inference nearly three times faster, with 200+ tokens per second in everyday use. A "Pro" model is said to arrive later.Source: minimax.io · checked 2026-11-17
- MiniMax’s API guide lists MiniMax-M3.1-Flash with a 1,000,000-token context window and text, image and video input. Thinking is on by default and cannot be turned off; the effort level can be low, medium, high, xhigh or max, and max is the default. The guide does not state a maximum output length.Source: platform.minimax.io · checked 2026-11-17
Hy4 Flash
Tencent$0.04 / $0.14 (per million tokens)
- Tencent’s Hy4 flash announcement, dated November 16, 2026, presents it as the newest member of the Hy model family. Tencent recommends it for most everyday tasks (deep research, document processing, PowerPoint generation, simple front-end work, computer use) and points complex tasks such as long-horizon software engineering to Hy4. Tencent Cloud’s API guide calls it "a leaner, more efficient everyday workhorse".Source: tencent.com · checked 2026-11-17
- The Tencent announcement is dated November 16, 2026, and Tencent Cloud’s model, pricing and API pages for hy4-flash were last updated the same day. The announcement says only that Hy4 flash is available now; it gives no separate release date or version name.Source: tencent.com · checked 2026-11-17
- Tencent says Hy4 flash has 224B total parameters (including 64B Engram parameters), 12B active parameters, a context of up to 1M tokens and multimodal input; the announcement does not state a maximum output or the output types. It says the KV cache is 1/6 the size of Hy4’s and that decoding exceeds 150 tokens per second. These are the maker’s own figures, not tested by us.Source: tencent.com · checked 2026-11-17
FAQ
MAI Code 2 Flash FAQ
What is MAI Code 2 Flash?
MAI Code 2 Flash is the lightweight model of Microsoft AI’s MAI Code 2 family, announced on November 17, 2026 together with MAI Code 2. Microsoft says it is built for high-volume work such as code explanations, targeted edits, routine fixes and rapid iterations, and GitHub calls it a cost-efficient model for everyday development workflows. It is meant for agentic coding in GitHub Copilot.
How much does MAI Code 2 Flash cost?
In GitHub Copilot it is $0.05 per million input tokens, $0.005 per million cached input tokens and $0.25 per million output tokens until May 17, 2027, then $0.10, $0.01 and $0.50 (1 GitHub AI credit is $0.01). Microsoft’s own cost claims use the promotional price, so they will look worse after May 17, 2027. We found no price for Microsoft Foundry.
Who can use MAI Code 2 Flash, and is it on the Copilot Free plan?
GitHub lists it for Copilot Free, Student, Pro, Pro+, Max, Business and Enterprise, to eligible subscribers on annual plans, in the model picker of VS Code, Visual Studio, Copilot CLI, the cloud agent, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse. MAI Code 2 itself is not listed for Free or Student. In Microsoft Foundry, Flash is in limited-access preview.
Is MAI Code 2 Flash better than Claude Haiku 5.5?
Microsoft says Flash offers comparable performance to Claude Haiku 5.5 at 20% of the cost, and its model page adds 12% higher performance; these are the maker’s own figures, not tested by us, and no Flash scores are published on the pages we read. On GitHub’s table the promotional price is half of Haiku’s. See Claude Haiku 5.5 for the model you can run here.
What are the limits of MAI Code 2 Flash, and what is not stated?
Microsoft says the new models are explicitly not yet tuned for harder problems like fundamental science research and very long-horizon agentic work. The pages we read state no context window, maximum output, API model id or image input for Flash, and Microsoft’s own Flash page returned "not found". Check Microsoft’s model card before you depend on any of these.
How is MAI Code 2 Flash different from MAI Code 2 and MAI-Code-1.1-Flash?
MAI Code 2 is the stronger sibling at ten times Flash’s per-token price, for deep debugging and multi-file work; it is covered in MAI Code 2 vs Claude Opus 5.5. MAI-Code-1.1-Flash is the predecessor, priced at $0.20 / $1.20 on GitHub and deprecated in Copilot on December 18, 2026, when GitHub points users to the latest MAI Code Flash model.
Can I use MAI Code 2 Flash here or on OpenRouter?
No, not yet. It is not on OpenRouter’s model list as of November 19, 2026, so this site cannot run it, although Microsoft says OpenRouter is coming "soon" with no date. On this site a typical reply to Claude Haiku 5.5 costs about 1 credit (a credit is about $0.01 in the $5 pack), billed by its real token cost, at least 1 credit. MiniMax M3.1 Flash and Hy4 Flash are other cheap options.
Is this the official Microsoft MAI website?
No. This page is unofficial and not affiliated with or endorsed by Microsoft. Facts link to the pages we read.