Agnes AI Launches Free, Uncapped Agentic Coding Model for Developers

Singapore-based AI startup Agnes AI has launched Agnes-2.5-Flash, the latest generation of its text model built specifically for agentic coding workflows, alongside a new desktop workspace called Agnes Code. The release is available immediately with free, uncapped access — no usage limits, no metered billing, and no announced end date.

The shift toward autonomous coding agents has been one of the defining trends in AI this year, with developers increasingly relying on models that can edit across multiple files, call tools, execute commands, and recover from errors without human intervention. But the deepening of this workflow has come with a well-documented cost: developers running agents daily are hitting usage caps and facing ballooning bills precisely when agentic coding becomes indispensable. Agnes-2.5-Flash directly targets this pain point by removing the cost barrier entirely rather than managing it.

Agnes-2.5-Flash is a direct successor to Agnes-2.0-Flash, which already counts more than 585,000 users via the company's API. The new model improves on code generation, project-context retention, tool-calling accuracy, and multi-turn task recovery. The underlying Agnes Harness orchestration layer has also been upgraded. A larger flagship model, Agnes-2.5-Pro, is set to follow later this month for teams tackling production-grade systems and complex agent pipelines.

Agnes Code, launching alongside the model, gives developers a project-based workspace where they can describe tasks in plain language and have them executed directly inside real codebases — reading file structure, editing across files, and running commands, with results saved to the project rather than scattered across a chat log. It runs on the same account and credits as Agnes AI's web and API products, so developers use a single balance across all surfaces.

Why it matters for Singapore: Agnes AI is one of Singapore's most prominent homegrown AI model companies, and the launch of a free, uncapped coding model at production scale signals the city-state's growing ambition in developer tooling. Agnes AI's infrastructure processed 5.41 trillion tokens in the past week alone — 3.25 trillion text and 2.16 trillion multimodal — under its existing free-access programme, making this release a meaningful step toward removing cost as a barrier to agentic development globally, from a Singapore base.