-
Notifications
You must be signed in to change notification settings - Fork 2.4k
feat: update kimi-k2-0905-preview and kimi-k2-turbo-preview #7663
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
feat: update kimi-k2-0905-preview and kimi-k2-turbo-preview #7663
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thank you for your contribution! I've reviewed the changes and found several issues that need attention before this can be merged.
| export type MoonshotModelId = keyof typeof moonshotModels | ||
|
|
||
| export const moonshotDefaultModelId: MoonshotModelId = "kimi-k2-0711-preview" | ||
| export const moonshotDefaultModelId: MoonshotModelId = "kimi-k2-0905-preview" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is the new kimi-k2-0905-preview model already available in production? Changing the default model could break functionality for users if this model isn't accessible yet. Have you verified its availability?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
| }, | ||
| "kimi-k2-0905-preview": { | ||
| maxTokens: 32_000, | ||
| contextWindow: 262_144, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The context window has been doubled from 131,072 to 262,144 tokens. Has this been tested with large contexts? This significant increase could impact:
- Memory usage and performance
- API rate limits
- Cost calculations for users
| outputPrice: 2.5, // $2.50 per million tokens | ||
| cacheWritesPrice: 0, // $0 per million tokens (cache miss) | ||
| cacheReadsPrice: 0.15, // $0.15 per million tokens (cache hit) | ||
| description: `Kimi K2 is a state-of-the-art mixture-of-experts (MoE) language model with 32 billion activated parameters and 1 trillion total parameters.`, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
The description is identical to the older kimi-k2-0711-preview model. If this is a newer version with improvements, could we update the description to reflect what's new or different about the 0905 version?
| "kimi-k2-turbo-preview": { | ||
| maxTokens: 32_000, | ||
| contextWindow: 131_072, | ||
| contextWindow: 262_144, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Similar concern here - the context window for kimi-k2-turbo-preview has been doubled. Has this change been tested for performance impact, especially given this is the 'turbo' variant optimized for speed?
|
The context widnow doesn't seem to match the official docs: https://platform.moonshot.ai/docs/introduction Am I missing something? |
|
Test is failing, but I can fix it in main. |
Related GitHub Issue
Closes: #
Roo Code Task Context (Optional)
Description
update kimi-k2-0905-preview and kimi-k2-turbo-preview
Test Procedure
Pre-Submission Checklist
Screenshots / Videos
Documentation Updates
Additional Notes
Get in Touch
Important
Update
moonshotDefaultModelIdand addkimi-k2-0905-previewwith updated context window and pricing.moonshotDefaultModelIdtokimi-k2-0905-previewinmoonshot.ts.kimi-k2-0905-previewwithcontextWindow262,144 and pricing details.contextWindowforkimi-k2-turbo-previewto 262,144.This description was created by
for 7c13a94. You can customize this summary. It will automatically update as commits are pushed.