Codex CLI: Memory Pipeline & Model Selection Fixes
CodexCodex CLI 0.101.0 was released on February 12, 2026, focusing on stability improvements to the model resolution system and the memory pipeline. The update ensures that model slug references remain stable during prefix-based selection, excludes developer messages from phase-1 memory input to reduce noise, and reduces memory processing concurrency to avoid race conditions during consolidation.
Sources & Mentions
2 external resources covering this update
Overview
Version 0.101.0 of the Codex CLI is a focused maintenance release published February 12, 2026. Rather than shipping new features, it addresses three specific stability issues identified in the memory and model-handling subsystems — areas that had been actively developed in the preceding 0.100.0 release.
Bug Fixes
Model Resolution Stability
A bug was corrected in which model slug references were being rewritten during prefix-based model selection. The fix ensures that when a user specifies a model by prefix (e.g., a shortened model name), the originally requested slug is preserved rather than substituted with an alternate identifier. This prevents unexpected model mismatches during sessions.
Developer Message Filtering in Memory
Developer messages — internal system-level messages not originating from the user — were previously being included in the phase-1 memory pipeline input. These messages introduced irrelevant or noisy content into the memory system. The fix excludes developer messages from phase-1 memory input, resulting in cleaner and more accurate memory consolidation.
Memory Concurrency Reduction
Memory phase processing was running with higher concurrency than was safe under load, which could produce instability during consolidation and staging operations. The concurrency level has been reduced to improve reliability during high-activity sessions.
Maintenance
The release also includes cleanup of phase-1 memory pipeline code paths and minor repository hygiene updates covering formatting and test-suite maintenance in remote model tests.