fix: Token limits for Amazon Bedrock Kimi K2 models

[msukmanowsky]

Summary

Corrects the context window and max output token limits for two Kimi K2 models on Amazon Bedrock:

• moonshot.kimi-k2-thinking
• moonshotai.kimi-k2.5

Changes

Field	Before	After
limit.context	256,000	262,143
limit.output	256,000	16,000

The previous values appeared to be placeholders. The corrected values match the actual limits advertised by Amazon Bedrock for these models.

https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html
https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-thinking.html

[rekram1-node]

mind linking doc to that?

[msukmanowsky]

@rekram1-node added links in the comments but also for your reference:

https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-5.html
https://docs.aws.amazon.com/bedrock/latest/userguide/model-card-moonshot-ai-kimi-k2-thinking.html

Neither actually specify the exact output limit so I've left it at 16k. Context window was validated empirically via manual testing.

[rekram1-node]

Thanks, and I think they do specify output limit see pic

[msukmanowsky]

@rekram1-node they do, but I was referring to the lack of specificity. I highly doubt it's actually 16,000 and more likely something like 16,384 (2**14). We may be a little conservative here, but better than the status quo where it assumed you can use the full context for output.

[msukmanowsky]

@rekram1-node bumping this!