alex/g3
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add GLM-4 to context length defaults (32k)

Browse Source 
GLM-4 models support 32k context but were falling back to the
conservative 4096 default, causing context overflow on startup.
This commit is contained in:
Dhanji R. Prasanna
2026-01-28 10:46:36 +11:00
parent e32c302023
commit 55dba121b7
1 changed files with 1 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

1
crates/g3-core/src/lib.rs
View File

@@ -663,6 +663,7 @@ impl<W: UiWriter> Agent<W> {
match &embedded_config.model_type.to_lowercase()[..] { match &embedded_config.model_type.to_lowercase()[..] {
"codellama" => 16384, // CodeLlama supports 16k context "codellama" => 16384, // CodeLlama supports 16k context
"llama" => 4096, // Base Llama models "llama" => 4096, // Base Llama models
"glm4" => 32768, // GLM-4 supports 32k context
"mistral" => 8192, // Mistral models "mistral" => 8192, // Mistral models
"qwen" => 32768, // Qwen2.5 supports 32k context "qwen" => 32768, // Qwen2.5 supports 32k context
_ => 4096, // Conservative default _ => 4096, // Conservative default