Use open-source models
with Claude Code
Connect Claude Code to Kyma API through the Anthropic-compatible Messages API. Great for when you need a second opinion or want to save costs.
Claude Alternative
When you hit Claude rate limits or want a second opinion, switch to open-source models instantly. Same CLI, different brain.
Save on High-Volume
Compare model prices for your tasks. Actual cost depends on the model, prompt length and output generated.
Active Models, One Key
Access Qwen, DeepSeek, Kimi, Gemma, Llama and more. All through one API key, one base URL.
Recommended Models
Top picks for Claude Code users: by quality, capability, and cost.
Supports reasoning, tool use, vision input. Compare current prices and measured uptime on the model page.
qwen-3.6-plusSupports reasoning, tool use, vision input. Compare current prices and measured uptime on the model page.
kimi-k2.6Supports reasoning, tool use. Compare current prices and measured uptime on the model page.
deepseek-r1Supports reasoning, tool use, vision input. Compare current prices and measured uptime on the model page.
gemma-4-31bSetup guide
1. Create account → 2. Set the env vars below → 3. Run Claude Code with any model.
# Connect to the Anthropic-compatible Messages API export ANTHROPIC_AUTH_TOKEN=kyma-xxxxx # your Kyma key export ANTHROPIC_BASE_URL=https://kymaapi.com # Then use Claude Code with --model flag claude --model qwen-3.6-plus # See the setup guide for persistent configuration.
# Keep Claude as primary, use Kyma models when needed: # For cost-sensitive tasks: claude --model gemma-4-31b "explain this codebase" # For reasoning tasks: claude --model deepseek-r1 "find the bug in this logic" # For agentic tasks: claude --model kimi-k2.6 "refactor auth module" # To restore your default connection, remove the Kyma environment variables.
More models, lower cost: pay per use
$0.50 signup credit for eligible free-tier models after email verification. No card required.
Create account