mirror of
https://github.com/QwenLM/qwen-code.git
synced 2025-12-19 09:33:53 +00:00
fix: token limits for qwen3-max (#724)
This commit is contained in:
@@ -115,7 +115,7 @@ const PATTERNS: Array<[RegExp, TokenCount]> = [
|
||||
[/^coder-model$/, LIMITS['1m']],
|
||||
|
||||
// Commercial Qwen3-Max-Preview: 256K token context
|
||||
[/^qwen3-max-preview(-.*)?$/, LIMITS['256k']], // catches "qwen3-max-preview" and date variants
|
||||
[/^qwen3-max(-preview)?(-.*)?$/, LIMITS['256k']], // catches "qwen3-max" or "qwen3-max-preview" and date variants
|
||||
|
||||
// Open-source Qwen3-Coder variants: 256K native
|
||||
[/^qwen3-coder-.*$/, LIMITS['256k']],
|
||||
@@ -179,8 +179,8 @@ const OUTPUT_PATTERNS: Array<[RegExp, TokenCount]> = [
|
||||
// Generic coder-model: same as qwen3-coder-plus (64K max output tokens)
|
||||
[/^coder-model$/, LIMITS['64k']],
|
||||
|
||||
// Qwen3-Max-Preview: 65,536 max output tokens
|
||||
[/^qwen3-max-preview(-.*)?$/, LIMITS['64k']],
|
||||
// Qwen3-Max: 65,536 max output tokens
|
||||
[/^qwen3-max(-preview)?(-.*)?$/, LIMITS['64k']],
|
||||
|
||||
// Qwen-VL-Max-Latest: 8,192 max output tokens
|
||||
[/^qwen-vl-max-latest$/, LIMITS['8k']],
|
||||
|
||||
Reference in New Issue
Block a user