Microsoft said its Code 2 coding model delivers 98% of OpenAI’s GPT-6.1 Sol performance at 40% of the price as it unveiled cheaper AI models at its Ignite conference in San Francisco on 17 November, Quartz reported. Microsoft says customers are choosing Copilot over Anthropic’s Claude and OpenAI’s ChatGPT because it costs less to operate, Bloomberg reported.
Key points
- Microsoft says Code 2 reaches 96% of Claude Opus 5.5’s performance at 5% of its cost.
- Code 2 and Code 2 Flash are generally available in GitHub Copilot and Foundry, according to Microsoft.
- The reported price comparisons use selected benchmarks, and promotional discounts account for part of the gap.
- Microsoft’s Copilot app also uses models from OpenAI and Anthropic.
Code 2 takes aim at Claude and GPT
Judson Althoff, chief executive officer of Microsoft’s commercial business, made the broader Copilot pitch during an Ignite keynote. “Anything you can do in Claude or ChatGPT, you can do better with Copilot, and for less,” he said, according to Bloomberg.
Against Anthropic’s Claude Opus 5.5, Microsoft says Code 2 achieves 96% of the performance at 5% of the cost. The company describes Code 2 as a family of coding models designed to act on software-development tasks. Chief Executive Satya Nadella announced it during the opening keynote alongside Code 2 Flash and Cyber 2 Flash.
Microsoft says Code 2 and Code 2 Flash are generally available through GitHub Copilot and Foundry, its AI model platform. The coding and cybersecurity models unveiled at Ignite give Microsoft products built in-house to offer alongside models supplied by other developers. Its Copilot app continues to use OpenAI and Anthropic models as well as Microsoft’s own, Quartz reported.
Code 2 Flash and Cyber 2 Flash prices
Microsoft positions Code 2 Flash for frequent, shorter coding jobs, including explanations, focused edits and routine fixes. It says the model matches OpenAI’s GPT-6 Luna at 65% of its price. Against Anthropic’s Claude Haiku 5.5, Microsoft claims comparable performance at one-fifth of the price.
For security work, Microsoft says Cyber 2 Flash beats OpenAI’s GPT-6 Astra on a security benchmark while costing less than 2% as much per task. The company is integrating Cyber 2 Flash with Perception, its agentic security system, which Nadella said became generally available on 17 November.
The comparisons rely on a hand-picked set of benchmarks, and promotional discounts account for some of the reported price differences, Quartz reported. Microsoft AI chief Mustafa Suleyman said tailoring models to particular domains lets the company fit more knowledge into the same number of parameters. A model focused on routine coding work handles complex mathematics or scientific research less well.
Microsoft expands its own model range
Microsoft introduced six specialised models as it works towards building its own general-purpose frontier model, Microsoft AI chief Mustafa Suleyman told Axios. The company has long used OpenAI and other developers for its most advanced models. At its Build conference in June, Microsoft introduced seven in-house MAI models and framed the move as part of a longer-term effort to rely less on outside providers.
Microsoft also announced MAI Licensed Weights, a programme through which it plans to give trusted customers access to its model weights. Those customers must deploy the licensed models within Azure, according to Quartz.