[FEATURE]: Add Kimi K2 models to Groq provider #9135

Description

Groq now offers Kimi K2 models through GroqCloud, but they are not currently available in OpenCode's Groq provider integration.

Current Status

Groq has recently added support for Kimi K2 models:

• Model ID: moonshotai/kimi-k2-instruct-0905
• Provider: Groq (GroqCloud)
• Pricing: $1.00 input / $3.00 output per 1M tokens
• Rate Limits: 250K TPM / 1K RPM (Developer Plan)
• Context Window: 131,072 tokens
• Max Completion Tokens: 8,192

References:

• Groq Blog: https://groq.com/blog/introducing-kimi-k2-0905-on-groqcloud
• GroqCloud Docs: https://console.groq.com/docs/model/moonshotai/kimi-k2-instruct-0905
• Groq Models Page: https://console.groq.com/docs/models

Expected Behavior

Users should be able to use Kimi K2 models through the Groq provider in OpenCode:

{
  "model": "groq/kimi-k2-instruct-0905"
}

Or in agent configuration:

{
  "agent": {
    "my-agent": {
      "model": "groq/kimi-k2-instruct-0905"
    }
  }
}

Why This Matters

• Kimi K2 is a high-quality MoE model with 1 trillion total parameters
• Now available through Groq's fast inference infrastructure
• Provides users with more model options when using Groq provider
• Maintains parity with Groq's official offerings

Implementation Notes

The model should be added to the Groq provider configuration in the models.dev API or OpenCode's provider definitions, similar to how other Groq models are currently configured.

Related

• Groq provider currently supports: Llama 3.1, Llama 3.3, Llama Guard, GPT-OSS, Whisper, etc.
• Kimi K2 is also available through Moonshot AI provider (moonshotai/kimi-k2-*)
• This adds another way for users to access Kimi K2 models