What changed
New default model: claude-opus-4-7[1m]. The [1m] suffix is NEXO’s internal notation that translates to the anthropic-beta: context-1m-2025-08-07 header at call time. The header is accepted by Opus 4.7, so the 1M context window continues to work exactly as before.
Reasoning effort set to max. Opus 4.7 introduced a new effort tier above xhigh. The valid values are low, medium, high, xhigh, and max — all verified empirically against /v1/messages. NEXO now defaults to the highest tier.
Auto-migration on nexo update. The heal_runtime_profiles() function detects users whose model starts with claude-opus-4-6 and silently migrates to claude-opus-4-7, preserving any suffix (e.g. [1m]). If the user had an empty reasoning effort, xhigh, or the legacy enabled format, it bumps to max.
Codex untouched. GPT-5.4 with xhigh reasoning remains the Codex default. No migration applies to Codex profiles.
Breaking API change in Opus 4.7
Opus 4.7 changed the thinking/reasoning API from thinking.enabled + thinking.budget_tokens to thinking.type: "adaptive" + output_config.effort. NEXO Brain does not call the Anthropic API directly (Claude Code handles it), but this is documented here for operators building custom integrations.
How to upgrade
Run nexo update in your terminal. If you were on Opus 4.6, the migration happens automatically. No manual configuration needed.