Enterprise AI spending is emerging as a major cost center for corporate finance teams. Glean, valued at $7.2 billion, announced its AI assistant consumed 70 percent fewer tokens than Anthropic's Claude across more than 180 enterprise tasks, translating to an 81 percent reduction in per-task token costs.
Glean's platform recorded an average token cost of $0.58 per task versus Claude's $2.98 for identical workloads. The efficiency gap stems from two core mechanisms: a pre-indexed, permission-aware context graph that retrieves relevant enterprise data without repeatedly querying expensive language models, and automatic model routing that selects cheaper alternatives for lower-complexity tasks.
The context layer connects over 275 business applications and maintains live permission tracking across platforms. Model routing dynamically selects from more than 40 AI models, avoiding default deployment of premium frontier models for every query.
Tests conducted between May and July 2026 compared Glean's system against off-the-shelf Model Context Protocol tools used in Claude. Human graders preferred Glean's results 78 percent of the time across correctness, response quality and overall interaction. Independent reviewers flagged certain discrepancies in benchmark construction but did not dispute the core efficiency claim.
Glean's annual recurring revenue reached $300 million by May 2026, up from roughly $100 million 15 months prior. A Series F round in June 2025 valued the company at $7.2 billion.
The company also unveiled Tau, a desktop AI workspace offering enhanced model access and granular usage controls for IT administrators.
The benchmark targets Anthropic but Glean's primary competitors are Microsoft and Google, which integrate Copilot and Gemini into productivity suites—Teams, Outlook, Word and Google Workspace—that most enterprises already use. Glean's permission-aware architecture addresses the critical enterprise concern of data control, allowing AI access across 275 connected applications while enforcing robust permission management.
