4 new providers, live now
You can now bring your own key from four more providers and run debates on them directly.
- OpenAI - connect your own OpenAI key to run debates on GPT-5.6
- Grok (xAI) - bring an xAI key and debate with Grok 4.5
- AWS Bedrock - use your AWS credentials to run Claude models (Opus 4.8 and Sonnet 5) through Bedrock
- Mistral AI - connect a Mistral key to run debates on Mistral Large
Each one shows up in the provider list on the debate creation page and in Settings > Provider Keys, with its own setup guide and estimated cost range per debate. Keys are never stored on our servers - they stay in memory for the length of the request, same as every other BYOK provider we support.
Updated models on providers you already use
If you're already running debates through Groq, DeepSeek, or Kimi, you're now on newer model generations by default:
- Groq: moved from Llama 3.3/3.1 to GPT-OSS 120B and GPT-OSS 20B
- DeepSeek: moved from DeepSeek R1/V3 to DeepSeek V4 Pro and V4 Flash
- Kimi: moved from Kimi K2 to Kimi K3
Cost ranges have been updated to match the new model pricing on each provider's page.
Anthropic support brought current
We also updated the Anthropic SDK we use under the hood, so debates routed to Claude now run on current-generation models (Opus 5, Sonnet 5, and Haiku 4.5) instead of the previous generation.
Where to find it
Open the debate creation page or go to Settings > Provider Keys to see the full provider list, connect a key, and start a debate on any of the models above.