Development build of main (019c9c7). It is not the latest release: sushi update, the built-in update check and Homebrew stay on v1.6.0 and never offer it. Ad-hoc signed and not notarized, like every sushi release.
To try it:
shasum -a 256 -c sushi-bin-macos-arm64.tar.gz.sha256
tar -xzf sushi-bin-macos-arm64.tar.gz
xattr -dr com.apple.quarantine sushi-macos-arm64
./sushi-macos-arm64/sushi --versionSince v1.6.1-dev1: DeepSeek-V4.1-Flash on Sushi Cooler decodes about 3% faster and draws a sampled token with one GPU kernel (a seed still gives one text on every repeat, but not the text an earlier build gave for it). Chat templates now render in sushi's own Zig template engine, and chat requests and pack loads are sturdier: an OpenAI function-role message is read as a tool result, and a pack with malformed tokenizer files fails to load with a line that names the key.