1.73.4

• bfdc3fe: Add support for Claude Opus 4.8 (issue #1832). The bare opus alias for the claude tool now resolves to claude-opus-4-8, and the explicit opus-4-8 / claude-opus-4-8 aliases (plus their [1m] variants for the 1M-token context window) are accepted everywhere existing Opus aliases are. All earlier aliases keep working unchanged — opus-4-7 / claude-opus-4-7, opus-4-6 / claude-opus-4-6, opus-4-5, sonnet, haiku, and opusplan continue to map to the same model IDs as before. The --fallback-model default chain for the claude tool extends to opus/opus-4-8 → opus-4-7 → opus-4-6; the --think xhigh/max levels remain supported (4.8 shares Opus 4.7's effort surface and adaptive-only thinking, so Claude Code never emits MAX_THINKING_TOKENS for it); --show-thinking-content still opts into thinking output on 4.8 the same way it does on 4.7. Adds the deep case study under docs/case-studies/issue-1832/ (covering the requirements, solution plan, and verification matrix) and tests/test-opus-48-model-support.mjs (175 assertions) alongside the existing 4.7 regression test. The English docs/CONFIGURATION.md row text is left unchanged in this PR to keep all four language siblings in sync; the case study is the authoritative user-facing documentation for the 4.8 behavior.

Related Pull Request: #1833

---