Status line shows ~200k context remaining when using 1M context window model

Bug Description

When selecting Opus 4.6 with 1M context (claude-opus-4-6[1m]), the status line shows left:199.9k instead of reflecting the 1M token window.

Root Cause

This is in my custom status line script (statusline-command.sh). The window_tokens variable is hardcoded to 200000 for all models:

case "$raw_model" in
    *sonnet*) display_model="sonnet-4.6"; window_tokens=200000 ;;
    *opus*)   display_model="opus-4.6";   window_tokens=200000 ;;
    *haiku*)  display_model="haiku-4.5";  window_tokens=200000 ;;
    *)        display_model="$raw_model"; window_tokens=200000 ;;
esac

The left value is then calculated as:

remaining_tokens=$(( window_tokens - total_in ))

So it always computes remaining tokens against a 200k window, even when the model is using the 1M context window (claude-opus-4-6[1m]).

How Context Window Size Is Determined

The status line reads context data from Claude Code's stdin JSON via jq:

• context_window.used_percentage — percentage of context used
• context_window.total_input_tokens — total input tokens consumed
• context_window.total_output_tokens — total output tokens consumed

These fields come from the Claude Code harness (provided to the status line command via stdin). However, the max context window size is not provided as a field — it is inferred in the script by matching the model name. The script currently does not distinguish between 200k and 1M variants of the same model.

Purpose

The status line displays [ctx:X% | used:Xk | left:Xk] to give the user a live view of how much context window capacity remains, so they know when they're approaching the limit and context compression will kick in.

Expected Fix

The model name matching should detect the 1M context window variant (e.g., model ID containing [1m] or 1m) and set window_tokens=1000000 accordingly.

Environment

• Claude Code version: 2.1.80
• OS: macOS 26.3 (arm64)