diff --git a/mappings/openinfer/gemma-4-31b.yaml b/mappings/openinfer/gemma-4-31b.yaml new file mode 100644 index 0000000..23bca2d --- /dev/null +++ b/mappings/openinfer/gemma-4-31b.yaml @@ -0,0 +1,15 @@ +model_id: gemma-4-31b +provider_id: openinfer +provider_model_name: "@oi/Gemma-4-31B-It" +pricing: + input_per_million: 0.52 + output_per_million: 0.75 + currency: USD +rate_limits: + requests_per_minute: 60 + tokens_per_minute: 100000 +context_window_override: 8192 +available_regions: + - global +created_at: "2026-09-08T00:00:00.000Z" +updated_at: "2026-09-08T00:00:00.000Z" diff --git a/mappings/openinfer/llama-3-2-1b.yaml b/mappings/openinfer/llama-3-2-1b.yaml new file mode 100644 index 0000000..e1c3e71 --- /dev/null +++ b/mappings/openinfer/llama-3-2-1b.yaml @@ -0,0 +1,15 @@ +model_id: llama-3-2-1b +provider_id: openinfer +provider_model_name: "@oi/Llama-3.2-1B-Instruct" +pricing: + input_per_million: 0.02 + output_per_million: 0.02 + currency: USD +rate_limits: + requests_per_minute: 60 + tokens_per_minute: 100000 +context_window_override: 24000 +available_regions: + - global +created_at: "2026-08-19T00:00:00.000Z" +updated_at: "2026-09-08T00:00:00.000Z" diff --git a/mappings/openinfer/qwen3-5-27b.yaml b/mappings/openinfer/qwen3-5-27b.yaml new file mode 100644 index 0000000..8ca6606 --- /dev/null +++ b/mappings/openinfer/qwen3-5-27b.yaml @@ -0,0 +1,15 @@ +model_id: qwen3-5-27b +provider_id: openinfer +provider_model_name: "@oi/Qwen3.5-27B" +pricing: + input_per_million: 0.72 + output_per_million: 0.72 + currency: USD +rate_limits: + requests_per_minute: 60 + tokens_per_minute: 100000 +context_window_override: 32768 +available_regions: + - global +created_at: "2026-09-08T00:00:00.000Z" +updated_at: "2026-09-08T00:00:00.000Z" diff --git a/mappings/openinfer/qwen3-5-9b.yaml b/mappings/openinfer/qwen3-5-9b.yaml new file mode 100644 index 0000000..df712dc --- /dev/null +++ b/mappings/openinfer/qwen3-5-9b.yaml @@ -0,0 +1,15 @@ +model_id: qwen3-5-9b +provider_id: openinfer +provider_model_name: "@oi/Qwen3.5-9B" +pricing: + input_per_million: 0.15 + output_per_million: 0.18 + currency: USD +rate_limits: + requests_per_minute: 60 + tokens_per_minute: 100000 +context_window_override: 65536 +available_regions: + - global +created_at: "2026-08-19T00:00:00.000Z" +updated_at: "2026-09-08T00:00:00.000Z" diff --git a/models/llama-3-2-1b.yaml b/models/llama-3-2-1b.yaml new file mode 100644 index 0000000..a0d4966 --- /dev/null +++ b/models/llama-3-2-1b.yaml @@ -0,0 +1,13 @@ +id: llama-3-2-1b +name: Llama 3.2 1B Instruct +description: Meta's smallest open-weight model with 1 billion parameters from the Llama 3.2 family. Designed for on-device and edge deployment, supporting dialogue, summarization, rewriting, and instruction following with a 128K context window. +capabilities: + - chat + - completion +modalities: + - text +context_window: 131072 +licensing: other +country: us +created_at: "2026-08-19T00:00:00.000Z" +updated_at: "2026-08-19T00:00:00.000Z" diff --git a/models/qwen3-5-27b.yaml b/models/qwen3-5-27b.yaml new file mode 100644 index 0000000..9df7df6 --- /dev/null +++ b/models/qwen3-5-27b.yaml @@ -0,0 +1,20 @@ +id: qwen3-5-27b +name: Qwen 3.5 27B +description: Alibaba's mid-size 27 billion parameter multimodal model from the Qwen3.5 family. Shares the hybrid architecture of its 9B sibling, combining Gated Delta Networks and Gated Attention for a 262K native context window, with text, image, and video understanding and coverage of 201 languages. Trades throughput for stronger reasoning, agentic tool use, and coding performance. Released under Apache 2.0. +capabilities: + - chat + - completion + - function-calling + - vision + - code-generation + - reasoning +modalities: + - text + - image + - video + - code +context_window: 262144 +licensing: apache-2.0 +country: cn +created_at: "2026-09-08T00:00:00.000Z" +updated_at: "2026-09-08T00:00:00.000Z" diff --git a/models/qwen3-5-9b.yaml b/models/qwen3-5-9b.yaml new file mode 100644 index 0000000..f9cc9d6 --- /dev/null +++ b/models/qwen3-5-9b.yaml @@ -0,0 +1,20 @@ +id: qwen3-5-9b +name: Qwen 3.5 9B +description: Alibaba's compact 9 billion parameter multimodal model from the Qwen3.5 family. Hybrid architecture combining Gated Delta Networks and Gated Attention delivers a 262K native context window, text, image, and video understanding, and coverage of 201 languages. Strong reasoning and coding performance for its size, released under Apache 2.0. +capabilities: + - chat + - completion + - function-calling + - vision + - code-generation + - reasoning +modalities: + - text + - image + - video + - code +context_window: 262144 +licensing: apache-2.0 +country: cn +created_at: "2026-08-19T00:00:00.000Z" +updated_at: "2026-08-19T00:00:00.000Z" diff --git a/providers/openinfer.yaml b/providers/openinfer.yaml new file mode 100644 index 0000000..66e1035 --- /dev/null +++ b/providers/openinfer.yaml @@ -0,0 +1,14 @@ +id: openinfer +name: OpenInfer +description: >- + Hosted, OpenAI-compatible inference API serving popular open-source + models on OpenInfer OS. Routes every request across distributed + CPU, GPU, and NPU compute with SLA-aware routing and automatic + fallback for lower cost per token. +api_base_url: https://api.openinfer.ai/v1 +auth_type: bearer +regions: + - global +compatibility: openai +created_at: "2026-08-19T00:00:00.000Z" +updated_at: "2026-08-19T00:00:00.000Z"