At its Ignite conference in San Francisco on Tuesday, Microsoft announced a number of new AI models, including two new coding models.
Imaginatively named Code 2 and Code 2 Flash, these new models promise to come very close to models like GPT-6.1 Sol and Claude Opus 5.5 at less than half the cost.
As Microsoft notes, “MAI Code 2 is at the frontier of price-performance” for enterprise coding tasks. “It does all of this while also delivering changes code maintainers would merge, at a fraction of frontier cost.”
The promise here is that Code 2 gets to 98% of the performance of GPT-6.1 Sol at 40% of the cost and 96% of Claude Opus 5.5 at 5% of the cost (which also goes to show how expensive Anthropic’s models are to run right now).
The model, Microsoft says, has also been tuned to create “great dashboards and visualizations” and that it “excels at visual coding.”
At least for the OpenAI comparison, it’s worth noting that OpenAI today released GPT-6.2 Astra and Sol, with improved performance. That softens Microsoft’s claims a bit, but it’s worth noting that OpenAI, too, notes that at least on Agents’ Last Exam, GPT-6.2 Astra is now 5.5 points “ahead of Claude Fable 5.2 at just 5% of the cost” (which also goes to show how expensive Anthropic’s models are right now).

As for MAI Code 2 Flash, Microsoft argues that it can match Anthropic’s Claude Haiku 5.5 and GPT-6 Luna on benchmarks like DeepSWE 1.1, Terminal-Bench 2.1, and SWE Atlas at 20% and 65% of the cost, respectively.
One interesting wrinkle here is that Microsoft specifically tuned these models for the day-to-day coding work of its users. “They’re explicitly not yet tuned for harder problems, like fundamental science research and very long horizon agentic work,” Microsoft writes.
The models Microsoft is comparing these models to are generalists. For a while, OpenAI took a similar approach with its Codex models, but it has abandoned this approach in favor of more general models.
Microsoft, meanwhile, which doesn’t build those kinds of in-house frontier models, seems to be focused on smaller specialized models that it can run more efficiently. And while they won’t match the frontier models across every benchmark, focusing on these specific tasks makes a lot of sense in this context.
Microsoft is taking this even further with models like MAI Work Flash, a model trained to work across Microsoft 365 and Fabric IQ products.

MAI Code 2 and Code 2 Flash are now available in GitHub Copilot, while Code 2 Flash is in limited preview on Microsoft Foundry. OpenRouter, Vercel, Hermes, Open Code, and Kilo will soon get support as well.
The post Microsoft’s new coding model promises near-Opus performance at 5% of the cost appeared first on The New Stack.
Generally available. Requires Microsoft 365 license.
An approved extension may still need boundaries. Only on Edge for Business, extension leak control prevents organization-installed extensions from reading content on protected pages. Those extensions can still work on other pages.
https://www.youtube.com/watch?v=zVRkf_640QM
Available in preview. Requires Microsoft 365 E3 licensing.
Available in preview. Requires Microsoft Sentinel licensing.
Available in preview with no additional licensing required.
We’re also making device security easier to verify. Windows baseline protection, now in preview in the Windows Insider Program, brings foundational protections like secure boot, virtualization-based security and memory integrity together into one verifiable security state. People get better visibility into the security posture of their PC, and IT gets a clear yes-or-no answer for every device in the fleet. Application developers can use baseline protection to verify that the devices running their applications meet their security requirements. The baseline will continue to evolve as we raise the bar in security.
We are proud to be collaborating with partners from across the ecosystem to see how Windows baseline protection helps meet their needs.
We’re also making it easier to protect your devices from the risks that come with elevated privilege. With Windows administrator protection, now generally available, IT can enable local administrators to run with standard user privileges by default. When a task needs administrator rights, Windows Hello verifies the user and grants those rights to that task alone. When the task ends, so do the elevated privileges, and the risk that came with them.
https://www.youtube.com/watch?v=lGpmGHrrZNg
The same AI that’s helping organizations scale is also helping attackers move faster. Our latest Microsoft Digital Defense Report found that weaponizing a vulnerability, which once took expert skills, now takes just one prompt. New vulnerabilities across the industry are on track to double this year. We’re using AI to stay ahead. MDASH, Microsoft Security’s multi-model agentic scanning harness, now scans Windows code at a scale we’ve never had before, finding vulnerabilities earlier and fixing them faster. In October alone, we shipped fixes for more than 1,100 vulnerabilities, more than in all of 2025. Those fixes only protect your employees once they reach your devices.
Our fastest customers deploy security updates to more than two-thirds of their devices within a week. Windows Autopatch helps more organizations move at that speed: on average, Autopatch customers deploy updates over 40% faster than customers using on-premises management, and usage has grown more than 50% in the last year, to over 150 million enrolled devices. Autopatch now also gives IT more control over updates, better visibility into the features they include and fewer interruptions, with one reboot a month.
The same management principles now apply to apps. With Microsoft Intune Enterprise App Management, IT can deploy and manage apps at scale and can now preview automatic updates not just for the applications they deploy, but the applications people install themselves, with the same controls and deadlines they know from Windows Autopatch.
Windows is also standardizing on declarative management, so devices keep themselves in the state IT defines. Policies such as Wi-Fi and VPN apply, within minutes and stay protected from drift, even when a device is offline.
All of this builds on the improvements we’ve made to Windows quality throughout the year, across performance, reliability and craft.
Learn more about Windows security and management at Ignite.
Leading agent builders and customers are integrating with MXC. OpenAI has turned MXC on by default in Codex, Meta is running MXC on tens of thousands of Windows devices, and more than 20 agent and app builders, including Anthropic, Box, Perplexity and Replit, are building on MXC.
https://www.youtube.com/watch?v=FXvmLQVadwg
Agent 365 governance and policies extend to local agents, including on Windows, so together with Microsoft Entra, Intune, Defender and Purview, organizations can govern agents with the practices they already use.
Learn more about Windows security, management, and agent governance at Ignite.
“Adobe’s flagship Windows apps like Premiere and After Effects already use local AI to deliver creative value to our users. Like many leading organizations, we’re looking to optimize growing AI spend, both ours and our customers’, without giving up access to frontier capabilities. We’re excited for more ways to harness hybrid intelligence to empower our developers to run key workflows in the right place, whether that’s local or cloud.”
— Ely Greenfield, CTO and EVP of Adobe’s Creative Products
Agent builders like Hermes, OpenClaw, Perplexity, Replit and more are building local AI experiences on Windows. Christie’s, the fine art and luxury auction house, uses Replit’s agent on Windows to work with private art collection data that never leaves the device. https://www.youtube.com/watch?v=0fjfAhKT0sA
Partners collaborating with the Windows endpoint security platform[/caption]
Alongside the Driver Quality Initiative and broader work to reduce kernel exposure and strengthen recovery, these investments are helping make Windows more resilient by design.
See what's changing with Windows endpoint security platform GA, driver quality and ecosystem resilience
We also know how important it is for users and organizations to better understand the security state of their devices. With Windows baseline protection, now available in preview, Windows brings foundational protections, including secure boot, virtualization-based security, and memory integrity together into a single, versioned security state. Instead of requiring customers and applications to understand and check individual protections, baseline protection makes it easier for users and IT to understand whether a device is in a measurable security state, to keep it up to date with the latest security safeguards, and, when stronger assurance is required, verify that state through hardware-backed attestation. It also provides developers a way to ensure that a Windows device running their applications is operating in a secure state. Windows baseline protection will continue to evolve as more security capabilities are added.
[caption id="attachment_180591" align="alignnone" width="1024"]
Windows baseline protection as part of the Windows settings experience[/caption]
Baseline protection will show up in an updated Windows Settings experience, making security and privacy protections easier for people to understand.
Learn how Windows is making its security foundation easier to establish, understand and verify with baseline protection
A trusted foundation must also prepare for threats that emerge before they become practical at scale. At Ignite, Windows is expanding post-quantum protection in the Windows TLS stack, including pure post-quantum key exchange and authentication, while advancing a broader crypto-agile foundation across the platform. These investments give organizations practical steps they can begin taking now while preparing for a multi-year transition across applications, infrastructure and identity.
Explore the new post-quantum TLS capabilities and how to begin preparing
Together, these investments help organizations establish a known and measurable security state, build on a more resilient platform, and prepare that foundation for emerging threats.
Windows Hello Face and PIN sign in available at Windows sign-in[/caption]
Administrator protection helps users operate with standard permissions by default and elevate privileges when needed, reducing the risk associated with standing administrator access.
Dive into passwordless, hardware-backed credentials and expanded authentication scenarios
Keeping a device secure is an ongoing process, not a one-time configuration. Investments across Windows Autopatch, declarative management, and recovery help IT maintain intended state, respond to risk, and restore operations after disruption.
See how Windows management and recovery capabilities help organizations stay secure through change
We're also advancing how organizations manage the broader Windows estate. New Intune capabilities simplify application deployment, inventory, and lifecycle management, helping applications remain current and governed alongside the devices where they run.
Microsoft is also extending Intune and Windows Autopatch to eligible Windows Server scenarios, in preview, helping endpoint teams manage configuration, compliance, applications, updates, and security across Windows clients and supported servers. This extends consistent management and security across more of the Windows estate, while complementing Azure Arc for cloud-native server management.
Learn more about Microsoft Intune’s new capabilities across Windows Server, app management and more