Codex and GPT-5.4 compatibility moves forward in 2026.4.14

OpenClaw Daily Brief

1. Operator signal: codex and GPT-5.4 compatibility moved forward in 2026.4.14: OpenClaw added forward compatibility for gpt-5.4-pro and updated Codex model pricing/limits visibility ahead of upstream catalog lag.
2. Provider configuration breakage got a direct fix: The Codex catalog now includes apiKey in provider output, addressing a failure mode where custom models could be silently dropped from model catalogs.
3. Local-model reliability got concrete timeout and token accounting fixes: Ollama embedded-run timeout handling and usage reporting were tightened to reduce false compaction triggers and stream-cutoff confusion.
4. Upgrade hygiene improved for systemd and onboarding paths: openclaw doctor --repair now avoids re-embedding dotenv secrets into user systemd units, while onboarding probes were tuned to stricter endpoint requirements.
5. Action for operators today: If you run custom providers or mixed cloud/local models, run models list --probe and one end-to-end task after upgrade to validate catalog visibility, timeout behavior, and token accounting.

Got a tip? Send it to tips@clawnews.org