Skip to main content

Context Truncation?

Feature Request

Not sure where else to ask this. I did search and did not come up with any source.

While evaluating signing up for T3 Chat and asking Gemini about it’s lack of Memory capabilities (which I don’t mind at all), Gemini made this troubling assertion:

```

Context Window Management: To maintain its high-speed performance and flat-rate pricing model, T3 Chat enforces more aggressive context truncation on older messages within a single prolonged thread. Native interfaces typically allow for deeper conversation stacks before dropping historical tokens.

```

Is this true?! Upon searching for any posts/blog posts about this, I could not find any official/authoritative information. Please consider adding some information somewhere about this.

Status: Closed3 comments

Log in to comment and vote

Comments3

  • Mark Florkowski

    Team

    No that is false, we do not do that.

    It does not seem worthwhile to document every behavior that we don’t do

    • Pink Soup

      Context Window management is not just any/every behavior. It is a core aspect today that every tool and even provider is actively messing around with. Aspersions cast on your product around this are poison. Not reacting to poison, is wrong.

      • Mark Florkowski

        Team

        We have not published detailed documentation on context behavior because our focus has been product iteration rather than exhaustive docs, and there is not much precedent for consumer AI products documenting this level of implementation detail.

        I would consider adding such things in the future, but right now it is not something that we can/will be prioritizing.