# Live models

[Models](/models)

Compare live model IDs, endpoints, and token prices. Connect with one KeepBal key using the OpenAI-compatible SDK base URL https://keepbal.com/v1.

Charges add the applicable input, output, cache read, and cache write token amounts at their respective rates. Cached input is charged at its cache rate instead of the regular input rate. Embeddings are billed for input tokens only.

- [claude-haiku-4.5](/models/claude-haiku-4.5) — Input: $1.25 / Output: $6.25 / Cache read: $0.13 / Cache write: $2.5 per 1M tokens — /v1/messages
- [claude-sonnet-4.5](/models/claude-sonnet-4.5) — Input: $3.75 / Output: $18.75 / Cache read: $0.38 / Cache write: $7.5 per 1M tokens — /v1/messages
- [deepseek-v3.2](/models/deepseek-v3.2) — Input: $0.336 / Output: $0.504 / Cache read: $0.26 / Cache write: $0 per 1M tokens — /v1/chat/completions
- [gpt-5.4](/models/gpt-5.4) — Input: $3.13 / Output: $18.75 / Cache read: $0.32 / Cache write: $0 per 1M tokens — /v1/chat/completions
- [gpt-5.4-mini](/models/gpt-5.4-mini) — Input: $0.94 / Output: $5.63 / Cache read: $0.1 / Cache write: $0 per 1M tokens — /v1/chat/completions
- [jina-embeddings-v3](/models/jina-embeddings-v3) — Input: $0.07 per 1M tokens — /v1/embeddings
- [qwen3-max](/models/qwen3-max) — Input: $1.44 / Output: $7.2 / Cache read: $0.9 / Cache write: $0 per 1M tokens — /v1/chat/completions

[Tools](/tools) · [MCP documentation](/docs/mcp)