# Gemini CLI ships model support updates and agent-protocol fixes

Google releases a stable coding-agent CLI update; llama.cpp changes FP16 dot-product accumulation. Retrospective brief for 21:00–21:59 UTC on 29 September 2026.

## Gemini CLI 0.62.0 adds model support and structured MCP call titles

Gemini CLI 0.62.0 includes changes for Gemini 3.8 Flash and 3.5 Flash Lite support, formats MCP tool-call titles as structured signatures, and emits a tool-call update before requesting permission in ACP mode. It also preserves OAuth refresh tokens during refresh and improves PTY cleanup. The release notes establish CLI support changes rather than a separate announcement of model availability. 

[google-gemini/gemini-cli v0.62.0 release notes](https://github.com/google-gemini/gemini-cli/releases/tag/v0.62.0)

## AVX512-FP16 dot products accumulate in FP32

The llama.cpp b11262 development build accumulates FP16 dot products in FP32 on the AVX512-FP16 path. This changes arithmetic in a CPU inference kernel. The release notes do not quantify its effect on end-to-end inference accuracy or throughput, and GitHub marks the build as a prerelease. 

[ggml-org/llama.cpp b11262 release notes](https://github.com/ggml-org/llama.cpp/releases/tag/b11262)