MAI-Code-1.1-Flash: Microsoft's Cheaper Vision Coding Model
MAI-Code-1.1-Flash is Microsoft's vision-capable Copilot coding model: 256K context, 138B/5B MoE, ~73% lower list price than mai-code-1-flash ($0.20/$0.02/$1.20).
At a Glance
- ID:
mai-code-1.1-flash - Org: Microsoft (MAI)
- Modalities: Text + image in; text out (vision-capable coding)
- Architecture: Transformer + sparse MoE; 138B total, 5B active
- Context: 256K tokens
- Knowledge / pretrain cutoff: December 2025 (model card)
- Training window: March 2026 - August 2026
- Surface: GitHub Copilot (GA rollout Aug 11, 2026)
- Predecessor:
mai-code-1-flash(Build 2026; retiring Sep 10, 2026)
Pricing (GitHub Copilot list)
| Model | Input / 1M | Cached input / 1M | Output / 1M | Premium request (annual) |
|---|---|---|---|---|
| mai-code-1.1-flash | $0.20 | $0.02 | $1.20 | 0.25x |
| mai-code-1-flash (retiring) | $0.75 | $0.075 | $4.50 | (prior list) |
Sources: GitHub Copilot models-and-pricing docs; GitHub changelog (73% lower list vs 1-Flash; 0.25x multiplier). Billed at provider list under usage-based billing.
Availability in Copilot
- Copilot Free / Student: via auto model selection
- Pro / Pro+ / Max / Business / Enterprise: manual picker + auto
- Surfaces: Copilot CLI, cloud agent, Copilot app, Chat on GitHub, VS Code, Visual Studio, GitHub Mobile, JetBrains, Eclipse, Xcode
- Business / Enterprise: admins must enable the MAI-Code-1.1-Flash policy (off by default)
Selected vendor evals (self-reported; Copilot harness)
From the MAI-Code-1.1-Flash model card (same VS Code / Copilot harness for compared models; not LLM Stats verified):
| Benchmark | MAI-Code-1.1-Flash | Tokens | MAI-Code-1-Flash | Tokens |
|---|---|---|---|---|
| SWE-Bench Verified | 72.6% | 8.6K | 71.6% | 10.8K |
| Terminal Bench 2.1 | 62.9% | 17.0K | 51.7% | 14.2K |
Launch post also cites (vendor): +22% Terminal-Bench 2.1 in Copilot CLI vs prior, +15% on .NET tasks, +4% code survival, +9% return visits; tokens stream 25% faster and use 25% fewer tokens per task.
Vision / web-app internal benches (model card; self-reported):
| Task | Pass rate | Tokens |
|---|---|---|
| Text2WebApp (internal) | 74.1% | 17.1K |
| ScreenShot2WebApp (internal) | 42.1% | 10.5K |
| Vision2Web Level3 | 11.5% | 15.1K |
What's New vs MAI-Code-1-Flash
- Native vision for image-grounded coding (screenshots to code).
- Stronger CLI / Terminal-Bench and .NET focus from developer feedback.
- Adaptive solution-length control (short on easy asks, more budget on hard ones).
- Roughly quarter-price list rates via training + serving efficiency (hundreds of thousands of RL environments in Copilot, per launch post; model card: >150k RL environments).
When to Use It
Good fit: Lightweight Copilot coding loops, repo Q&A, refactoring, tool-using agents, screenshot-to-code where cost and latency matter more than a frontier coding model.
Watch: Business/Enterprise needs the policy flip. Outputs still need review (model card known limitations). Model card notes Copilot CLI support was "planned for a later rollout" in the card text; the Aug 11 changelog already lists Copilot CLI in the picker, so prefer the changelog for current surfaces.
Caveats
- Pricing on the model card itself says "To be finalized"; use the live Copilot pricing table.
- Benchmarks are vendor / in-harness; label as self-reported.
- 1-Flash retirement: Sep 10, 2026 (GitHub coverage / related changelog).
Sources
- https://microsoft.ai/news/mai-code-1-1-flash-br-better-faster-at-a-quarter-of-the-cost/
- https://github.blog/changelog/2026-08-11-mai-code-1-1-flash-available-in-github-copilot/
- https://microsoft.ai/pdf/MAI-Code-1.1-Flash-Model-Card.PDF
- https://docs.github.com/copilot/reference/copilot-billing/models-and-pricing
Questions
Frequently Asked Questions
August 11, 2026, in GitHub Copilot.
- Copilot list: $0.20 input / $0.02 cached / $1.20 output per 1M tokens; annual plans use a 0.25x premium-request multiplier (~73% cheaper list than mai-code-1-flash).
Yes. Native vision: text + image in, text out (screenshot-to-code style workflows).
- Model card: 138B total / 5B active MoE; 256K context; pretrain cutoff December 2025.
September 10, 2026 (GitHub coverage / related changelog).
Free/Student via auto selection; Pro and above via picker + auto. Business/Enterprise admins must enable the MAI-Code-1.1-Flash policy (off by default).
Continue Reading