SAN FRANCISCO — Glean announced Aug. 26 that its AI assistant consumed 70 percent fewer tokens than Anthropic's Claude Cowork across more than 180 enterprise tasks, translating to an 81 percent reduction in per-task token costs.
Internal benchmarks showed Glean's platform averaged $0.58 per task in token costs compared to Claude Cowork's $2.98 for identical workloads. The gap exposes a central pressure in enterprise AI deployment: as companies scale AI assistants across thousands of employees, token economics become a line-item problem that executive teams watch closely.
Glean attributes the savings to two architectural choices. First, a pre-indexed, permission-aware context graph that holds enterprise data and user access controls in a live layer, eliminating the need to repeatedly query expensive large language models from scratch. The system connects over 275 business applications. Second, model routing that automatically directs requests across more than 40 different AI models, selecting cheaper alternatives for tasks that don't require premium frontier models.
This is not a philosophical efficiency gain. For a 1,000-person company running AI agents across customer support, internal search, and document processing, the difference between $0.58 and $2.98 per task compounds into millions of dollars annually in cloud spend.
In separate evaluations conducted May to July 2026, human graders preferred Glean's outputs 78 percent of the time on correctness and response quality. Glean commissioned the evaluation; independent reviewers flagged methodology discrepancies, though the core token efficiency claim remained unchallenged.
Glean reached $300 million in annual recurring revenue by May 2026, up from roughly $100 million in February 2025. A Series F round closed in June 2025 valued the company at $7.2 billion.
The company also unveiled Tau, a desktop AI workspace offering tighter usage controls and IT administrator visibility into model access and data permissions—a direct response to enterprise security requirements.
Glean's competitive position, however, is more complex than a simple price comparison. Microsoft and Google embed AI assistants directly into Office and Workspace, making switching costs high regardless of per-token math. Anthropic, for its part, can trade margin for market share in enterprise deals. Whether cost leadership translates to durable competitive advantage depends on whether enterprises view AI as a commodity input—where token economics dominate—or as a feature bundled into larger productivity suites where convenience and integration matter more.
