diff --git a/providers/coralbricks/logo.svg b/providers/coralbricks/logo.svg new file mode 100644 index 0000000000..4677066155 --- /dev/null +++ b/providers/coralbricks/logo.svg @@ -0,0 +1,6 @@ + + + + + + diff --git a/providers/coralbricks/models/glm-5.2-fp4.toml b/providers/coralbricks/models/glm-5.2-fp4.toml new file mode 100644 index 0000000000..0ef1e241db --- /dev/null +++ b/providers/coralbricks/models/glm-5.2-fp4.toml @@ -0,0 +1,14 @@ +base_model = "zhipuai/glm-5.2" +name = "GLM 5.2 FP4" + +reasoning_options = [] +interleaved = true + +[limit] +context = 1_048_576 +output = 131_072 + +[cost] +input = 1.12 +output = 4.4 +cache_read = 0 diff --git a/providers/coralbricks/models/gpt-oss-120b.toml b/providers/coralbricks/models/gpt-oss-120b.toml new file mode 100644 index 0000000000..ae4e47fc62 --- /dev/null +++ b/providers/coralbricks/models/gpt-oss-120b.toml @@ -0,0 +1,9 @@ +base_model = "openai/gpt-oss-120b" + +reasoning_options = [] +interleaved = true + +[cost] +input = 0.12 +output = 0.6 +cache_read = 0 diff --git a/providers/coralbricks/models/kimi-k3.toml b/providers/coralbricks/models/kimi-k3.toml new file mode 100644 index 0000000000..3bc0fc3063 --- /dev/null +++ b/providers/coralbricks/models/kimi-k3.toml @@ -0,0 +1,9 @@ +base_model = "moonshotai/kimi-k3" + +reasoning_options = [] +interleaved = true + +[cost] +input = 3.0 +output = 15.0 +cache_read = 0 diff --git a/providers/coralbricks/provider.toml b/providers/coralbricks/provider.toml new file mode 100644 index 0000000000..bcd1d1f321 --- /dev/null +++ b/providers/coralbricks/provider.toml @@ -0,0 +1,17 @@ +# Reasoning HTTP format (accessed 2026-08-04): +# OpenAI Chat: POST https://inference.coralbricks.ai/v1/chat/completions. +# Reasoning models stream interleaved thinking via a `reasoning` delta field +# (OpenRouter-style). Request-side reasoning control is NOT declared: the +# gateway serves each request from local capacity or a vendor relay, and the +# `reasoning` request field (enabled/effort) is honored only on the relay path +# while local serving rejects it (400 "Extra inputs are not permitted"; +# verified 2026-08-04 against both paths). No toggle/effort until the gateway +# normalizes this. Token budget: not documented. Streaming always returns +# usage in the final chunk. +# Sources: +# https://www.coralbricks.ai/docs +name = "CoralBricks" +env = ["CORAL_API_KEY"] +npm = "@ai-sdk/openai-compatible" +doc = "https://www.coralbricks.ai/docs" +api = "https://inference.coralbricks.ai/v1"