Glean
Searches and answers questions across all your work apps
- Made by
- Glean
- Category
- Search
- Last change
- Oct 26, 2026
Changelog
October 2026
- Availability
Glean is now available on the Claude Marketplace, so Anthropic customers can apply part of their committed Anthropic spend to Glean.
What this means for you: If your company already commits to spending a set amount with Anthropic, you can now put part of that money toward Glean. Glean searches and answers questions across your work apps, bringing company context, answers and action to your team.
How to use it: Glean is listed on the Claude Marketplace. Anthropic customers can apply a portion of their committed Anthropic spend to it.
- New feature
Glean introduced headless agent building, letting developers scaffold agents with a prompt, keep specs in version control, and run evals from their coding environment.
What this means for you: Developers can now build Glean agents from inside their coding environment. They describe the agent in a prompt (using /glean_run) and Glean scaffolds it, which means it sets up the starting structure. They can keep the agent's specs in version control (a saved history of changes) and run evals (tests that check how well the agent works) there too.
- New model
Glean added Gemini Argon to its model hub, a model built for complex tasks that uses more tokens and has higher latency.
What this means for you: Glean has a new model choice, Gemini Argon, built for complex tasks. It uses more tokens (the small chunks of text AI reads and writes, which affect cost) and takes longer to answer than other models, so it suits harder jobs rather than quick lookups.
Why it matters: Glean's own tests found Gemini Argon preferred 61.5% of the time over GPT-Sol 6.1, and nearly as good as Claude Opus 5.5 at finishing tasks.
How to use it: Open Glean and pick Gemini Argon from its model hub.
My company uses Glean, which now offers Gemini Argon for complex tasks, though it uses more tokens and responds more slowly. Help me sort my usual Glean requests into ones worth running on a slower, heavier model and ones that should stay on a quick model. My typical tasks are: [list tasks].
- New feature
Glean added skill reflection, which learns from how skills are used and suggests improvements for users to review and approve.
What this means for you: Glean's skill reflection watches how your AI skills (saved routines for repeat tasks) get used, and learns from repeated feedback and patterns. It then suggests improvements, and you review and approve each one before it takes effect.
Why it matters: It follows Glean's Oct 13 launch of team chat in open beta, where teams create and review work together.
My team uses a saved AI routine for [task, e.g. weekly status summaries]. I just learned Glean can suggest improvements to skills based on repeated feedback and usage patterns, which people then approve. Help me list the feedback and corrections I should track, and how to review a suggested change before approving it. Here is how the routine works today: [describe steps].
- New feature
Glean launched team chat in open beta, a shared chat where teams create and review work together, which admins can turn on.
What this means for you: Team chat is a shared chat in Glean where your whole team works in one conversation with the same context. You can create, discuss and review work together instead of each person chatting alone with the AI.
How to use it: It is in open beta, meaning it is open to use but still being tested. Ask your admin to turn it on.
My team is about to try Glean's new team chat, a shared chat where people create, discuss and review work together. Help me plan how we should use it for a real project: who does what, what to put in the shared chat, and how to review results. Our team does: [describe team and project].
- New model
Glean added Claude Haiku 5.5, an efficient model for everyday tasks like finding information and generating structured output.
- New model
Glean added GPT Sol 6.1, which uses 57.8% fewer tokens per task than Sol 6 and costs a fifth of Astra-level models.
September 2026
- Availability
Glean joined OpenAI's new B2B marketplace, so eligible enterprises can put part of their existing OpenAI commitment toward Glean.
- New model
Glean added Claude Opus 5.5, GPT-6 Luna and GPT-6 Sol, each with lower token costs or prices than their predecessors.
- New model
Glean added OpenAI's GPT-6 Astra, which performs well on long-running tasks that build ready-to-use deliverables.
- Launch
Glean introduced Glean Transform, a map of how work gets done that helps teams find where AI adds value and measure impact.
- New model
Glean added Gemini 3.8 Flash, which completed more than 3x as many long, document-heavy tasks as Gemini 3.7 Flash in its evaluations.
- New model
Glean added Claude Fable 5.1, preferred roughly 2:1 over Fable 5 by its judges, with improved response grounding.
August 2026
- Launch
Glean introduced Glean AI Gateway, a governed layer for LLMs and MCP that gives enterprises control over models, context, skills and memories.
- Launch
Glean introduced Glean Intelligence, a new platform layer combining model choice, auto routing and AI usage tracking for enterprises.
- Launch
Glean launched Glean Tau, a desktop AI workspace that unifies local files, apps and code with enterprise context.
