Context-Bench also measures total cost to complete the benchmark. Surprisingly, raw token costs ($/million tokens) do not map directly to total cost.

GPT-5 has lower per-token cost than Sonnet 4.5, but costs more in the benchmark because GPT-5 agents are more "token hungry".

Today we're releasing Context-Bench, an open benchmark for agentic context engineering.

Context-Bench evaluates how well language models can chain file operations, trace entity relationships, and manage long-horizon multi-step tool calling.

Claude Haiku 4.5 is now available on Letta Cloud.

Agents that redesign their own memory architecture.

We just shipped the memory tool in Letta—agents can now autonomously create, modify, and delete their own memory blocks.

We built a self-improving agent from scratch to demo it in our new YouTube video.

youtu.be/0nfNDrRKSuU

Letta Cloud now provides Anthropic's `memory` tool, enabling agents to dynamically manage their own memory architecture. Agents can create, delete, view, and modify memory blocks as they learn.

In the video below, we use the memory tool in the ADE with Sonnet 4.5 to create a new memory block.

Archival memory is how Letta agents store arbitrary pieces of text they can recall for later use.

In this deep-dive video, you'll see how to build an AI agent designed for long-term memory: The Archivist.

Learn more: youtu.be/hFNWhrXukc0

Replacing Timber the dog with superintelligence

Did you know that you can copy any page on docs.letta.com as a markdown document?

It's true.

You should be able to find it in the top right of the ADE.

Where did you expect it to be? We could probably make it more prominent if you've got tips.

The work is mysterious and important.