Incident sheet · /enterprise-ai
- System
- Enterprise AI
- Filed
- Read
- 3 min · 685 words
- Severity
- Routine
Microsoft launches MAI-Code-2 at half GPT-6.1’s token price
Microsoft launched MAI-Code-2 at $1 per million input tokens, half GPT-6.1’s rate, while expanding coding, image, cyber and routing tools for enterprise Foundry and GitHub users.

Event log
- Microsoft launched MAI-Code-2 on Tuesday at Ignite 2026.
- MAI-Code-2 costs $1 per million input tokens and $5 per million output tokens.
- GPT-6.1 costs $2/$10 and Claude Opus 5.5 costs $4/$20 per million input/output tokens.
- Decision 1.1 uses Google DeepMind’s open-weight Gemma and supports routing, classification, prioritization and verification.
- Code-2-Flash is in private preview in Foundry; Image 2.7 and Cyber 2 Flash are in preview.
By Amara Osei
Microsoft launched MAI-Code-2 on Tuesday at Ignite 2026 with prices of $1 per million input tokens and $5 per million output tokens, half OpenAI’s listed GPT-6.1 rates.
The release forms part of a broader push into coding and security, two domains already served by OpenAI and Anthropic, while answering enterprise demand for lower costs and better workload routing.
After ending its exclusive OpenAI agreement in April, Microsoft now has to show that its own model and tooling stack can stand on its own.
What did Microsoft deploy?
Microsoft expanded the MAI family across several product areas:
- MAI-Code-2 handles demanding agentic coding, while MAI-Code-2-Flash targets high-volume, everyday coding. GitHub hosts both models.
- Microsoft has placed Code-2-Flash in private preview in Foundry, its platform for building and scaling AI applications and agents.
- MAI-Image-2.7 and MAI-Image-2.7-Flash entered preview for sharper image editing and more consistent generation.
- MAI-Cyber-2-Flash entered preview to help security teams find, confirm and fix vulnerabilities.
Microsoft also launched Decision 1.1, which uses Google DeepMind’s open-weight Gemma as its base. It supports fast routing that Microsoft calls cost-effective, along with classification, prioritization and verification. Developers can supply guidelines so the system makes the right decision for each AI workload.
Microsoft said trusted partners would get first access to MAI model weights. At Tuesday’s keynote, Chairman and CEO Satya Nadella said the company remained committed to ensuring that everyone could own their own intelligence.
Takeaway: Microsoft has widened its model catalogue, but several specialist releases still sit in preview.
How do the new prices compare?
MAI-Code-2’s standard price is $1 per million input tokens and $5 per million output tokens. GPT-6.1 costs $2 and $10 for the same units, while Anthropic’s Claude Opus 5.5 costs $4 and $20.
Code-2 therefore charges half as much as GPT-6.1 and one-quarter as much as Claude Opus 5.5 for both input and output.
Michael Heath, principal architect for cloud and AI at global systems integrator Myraid360 and a Microsoft partner, said: “With this release, the MAI family has matured into models that cost less and still hold up against the other big model makers.” Heath said Microsoft was competing on price-performance rather than raw intelligence.
Takeaway: Microsoft wins the published token-price comparison; workload performance remains the larger test.
What does automated routing change?
Foundry’s model router automatically selects the model for each use case. Decision 1.1 adds the decision layer around that routing, including classification, prioritization and verification.
Heath said the enterprise question has shifted from choosing a model to finding a platform that lets teams swap models without rebuilding their agents. Automatic routing, he said, lets customers concentrate on outcomes rather than choosing models or watching usage meters.
That lets customers change models without rebuilding the agents around them, turning model selection into a managed platform operation rather than a permanent application choice.
Takeaway: Routing turns model selection into a managed Foundry service instead of a fixed application dependency.
What will prove the platform works?
Mitch Ashley, an analyst at Futurum Group, set the harder test: “The question is whether its models give customers an advantage inside the products they already use.” He said Microsoft need not win the standalone model race if it delivers better economics and useful capabilities through GitHub Copilot, Foundry and its security portfolio.
Cyber-2 Flash offers a lower-cost, practical, specialized option for agent-driven security operations, but enterprises must validate its performance independently. A cheap model fails if it misses an attack or triggers the wrong remediation, and the same standard applies to Code-2 and Code-2-Flash.
Ashley said enterprises will judge the quality and total cost of completed work; token price is only one input to that calculation.
Takeaway: Microsoft can use lower token prices and easier routing to win trials, but adoption depends on the quality and total cost of completed work.
via techtarget.com (Source)
More from Amara Osei
Show full bio
Correspondent covering regulation, policy and their impact on enterprise information technology and the leaders who run at CIO Monitor.
187 articles
The newsroom digest
One email a day with the stories that matter. No noise.
Cross-references
- Microsoft Fires Back at OpenAI and Anthropic With MAI Code 2
- Microsoft launches MAI Code 2 models to cut reliance on OpenAI
- OpenAI Cuts Token Prices 20% With GPT-6.2 Astra and Sol Ahead of Anthropic IPO
- Microsoft's MAI models chase cheap tokens, not bragging rights
- Microsoft Unveils Maia 300, Targets 40% Higher Token Efficiency