| The EdStratum Labs x6 model is text-only. |
Images go to a vision model through describe_image; read_file returns an image’s OCR text. |
| The model’s own knowledge is dated. |
Use web search for current facts. |
| The model has no built-in web search. |
Search goes through a third-party provider. |
| Some requests, often with code or shell output, can be blocked by upstream filtering. |
X6 shows a short note with the request ID. |
| Input is limited to about 256k tokens. |
Long chats drop the oldest turns from what is sent; long runs shorten older tool results. |
| Restarting the server interrupts running tasks. |
Restart only when no run is active. |
