Skip to showcase
Organic LLM
Back to showcase

Inspect · 4 of 5

Context Controls

See what fills the next answer, then raise context effort and watch the memory budget grow.

Scripted demo · fictional data

Context effort

Heavy

Memory budget

6,000 tokens

Instant · Quick · Heavy: 400 · 2,500 · 6,000

Time allowed to retrieve

up to 5 s

Instant · Quick · Heavy: 250 ms · 1 s · 5 s

  • Plans its queries first (up to 1.2 s), then a second pass
  • Pulls 40 candidates, keeps the best 36
  • Profile: 6 sections, in depth

Context inspector

Breakdown for the next answer

11%in use

Next send estimate

13k / 119k input tokens

15% conversation · 85% scaffolding

In context

Thread in context
100%4 / 4 msgs
Memories
Not recorded
Tools armed
  • Memory

Capacity

Model
Sonar Reasoning Pro
Model window
127k tok
Free input space
106k tok
Headroom
~221 turns

Memories in context

4 of 4 in the window

  • Shoots on a full-frame mirrorless camera with a 24mm f/1.4 lens.
  • New to astrophotography — prefers settings explained step by step.
  • Based in San Francisco; keeps road trips under about four hours each way.
  • Gets cold at night — always wants an extra warm layer on packing lists.
Finished

Effort moves from Quick toward Heavy. Retrieval may wait longer, search wider, and carry more memory into the answer.

Scripted demo

Behind the interface. The context inspector separates the system prompt, messages, tools, and memories inside the model’s input window.