feat: add hand-owned model quirks overlay

quirks.yaml

@@ -0,0 +1,880 @@
+# Hand-owned model behavior overlay.
+#
+# Everything in this file is curated by hand and layered on top of API-sourced
+# model data: oauth blocks, request patches, wire-name aliases, behavioral
+# flags, and models that no public catalog provides.
+#
+# `models:` entries are self-complete — `rules` apply ONLY to API-sourced
+# models, never to hand-owned `models:` entries. Legacy verbatim-preserved
+# entries are grouped under a `# legacy — pruning candidates` comment.
+
+- provider: openai
+  rules:
+    - match: "gpt-5.6-*"
+      default_reasoning_effort: medium
+    - match: "gpt-5.5"
+      default_reasoning_effort: medium
+    - match: "gpt-5.5-pro"
+      default_reasoning_effort: high
+    - match: "gpt-5.4"
+      default_reasoning_effort: none
+    - match: "gpt-5.4-pro"
+      default_reasoning_effort: medium
+    - match: "gpt-5.4-mini"
+      default_reasoning_effort: none
+    - match: "gpt-5.4-nano"
+      default_reasoning_effort: none
+    - match: "gpt-5.3-codex"
+      default_reasoning_effort: medium
+    - match: "gpt-5.2"
+      default_reasoning_effort: none
+    - match: "gpt-5.2-pro"
+      default_reasoning_effort: medium
+    - match: "gpt-5.1"
+      default_reasoning_effort: none
+    - match: "gpt-5.1-chat-latest"
+      default_reasoning_effort: none
+    - match: "gpt-5"
+      default_reasoning_effort: medium
+    - match: "gpt-5-chat-latest"
+      default_reasoning_effort: medium
+    - match: "o4-mini"
+      system_prompt_prefix: Formatting re-enabled
+      default_reasoning_effort: medium
+      patch:
+        body:
+          max_tokens: null
+          temperature: null
+          top_p: null
+    - match: "o3"
+      system_prompt_prefix: Formatting re-enabled
+      patch:
+        body:
+          max_tokens: null
+          temperature: null
+          top_p: null
+    - match: "o3-mini"
+      system_prompt_prefix: Formatting re-enabled
+      patch:
+        body:
+          max_tokens: null
+          temperature: null
+          top_p: null
+  models:
+    - name: text-embedding-3-large
+      type: embedding
+      input_price: 0.13
+      max_tokens_per_chunk: 8191
+      default_chunk_size: 2000
+      max_batch_size: 100
+    - name: text-embedding-3-small
+      type: embedding
+      input_price: 0.02
+      max_tokens_per_chunk: 8191
+      default_chunk_size: 2000
+      max_batch_size: 100
+    # legacy — pruning candidates
+    - name: o4-mini-high
+      max_output_tokens: 100000
+      real_name: o4-mini
+      max_input_tokens: 200000
+      input_price: 1.1
+      output_price: 4.4
+      supports_vision: true
+      supports_function_calling: true
+      system_prompt_prefix: Formatting re-enabled
+      patch:
+        body:
+          reasoning_effort: high
+          max_tokens: null
+          temperature: null
+          top_p: null
+    - name: o3-high
+      real_name: o3
+      max_input_tokens: 200000
+      input_price: 2
+      output_price: 8
+      supports_vision: true
+      supports_function_calling: true
+      system_prompt_prefix: Formatting re-enabled
+      patch:
+        body:
+          reasoning_effort: high
+          max_tokens: null
+          temperature: null
+          top_p: null
+    - name: o3-mini-high
+      max_output_tokens: 100000
+      real_name: o3-mini
+      max_input_tokens: 200000
+      input_price: 1.1
+      output_price: 4.4
+      supports_vision: true
+      supports_function_calling: true
+      system_prompt_prefix: Formatting re-enabled
+      patch:
+        body:
+          reasoning_effort: high
+          max_tokens: null
+          temperature: null
+          top_p: null
+
+- provider: gemini
+  rules:
+    - match: "gemini-3.6-flash"
+      default_reasoning_effort: medium
+    - match: "gemini-3.5-flash"
+      default_reasoning_effort: medium
+    - match: "gemini-3-flash-preview"
+      default_reasoning_effort: high
+    - match: "gemini-3.1-flash-lite"
+      default_reasoning_effort: minimal
+    - match: "gemini-3.1-pro-preview"
+      default_reasoning_effort: high
+    - match: "gemini-3-pro-preview"
+      default_reasoning_effort: high
+    - match: "gemini-2.5-pro"
+      default_reasoning_effort: high
+  models:
+    - name: text-embedding-004
+      type: embedding
+      input_price: 0
+      max_tokens_per_chunk: 2048
+      default_chunk_size: 1500
+      max_batch_size: 100
+
+- provider: claude
+  rules:
+    - match: "claude-*"
+      require_max_tokens: true
+    - match: "claude-opus-5"
+      default_reasoning_effort: high
+    - match: "claude-fable-*"
+      default_reasoning_effort: high
+    - match: "claude-opus-4-8"
+      default_reasoning_effort: high
+    - match: "claude-opus-4-7"
+      default_reasoning_effort: high
+    - match: "claude-opus-4-6"
+      default_reasoning_effort: high
+    - match: "claude-sonnet-5"
+      default_reasoning_effort: high
+    - match: "claude-sonnet-4-6"
+      default_reasoning_effort: high
+  variants:
+    - match: "claude-opus-4-6"
+      suffix: ":thinking"
+      real_name: claude-opus-4-6
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-sonnet-4-6"
+      suffix: ":thinking"
+      real_name: claude-sonnet-4-6
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-sonnet-4-5-20250929"
+      suffix: ":thinking"
+      real_name: claude-sonnet-4-5-20250929
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-haiku-4-5-20251001"
+      suffix: ":thinking"
+      real_name: claude-haiku-4-5-20251001
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-opus-4-5-20251101"
+      suffix: ":thinking"
+      real_name: claude-opus-4-5-20251101
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-opus-4-1-20250805"
+      suffix: ":thinking"
+      real_name: claude-opus-4-1-20250805
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-opus-4-20250514"
+      suffix: ":thinking"
+      real_name: claude-opus-4-20250514
+      max_output_tokens: 24000
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+
+- provider: mistral
+  models:
+    - name: mistral-embed
+      type: embedding
+      max_input_tokens: 8092
+      input_price: 0.1
+      max_tokens_per_chunk: 8092
+      default_chunk_size: 2000
+
+- provider: ai21
+  models:
+    - name: jamba-large
+      max_input_tokens: 256000
+      input_price: 2
+      output_price: 8
+      supports_function_calling: true
+    - name: jamba-mini
+      max_input_tokens: 256000
+      input_price: 0.2
+      output_price: 0.4
+      supports_function_calling: true
+
+- provider: cohere
+  models:
+    - name: embed-v4.0
+      type: embedding
+      input_price: 0.12
+      max_tokens_per_chunk: 2048
+      default_chunk_size: 2000
+      max_batch_size: 96
+    - name: embed-english-v3.0
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+    - name: embed-multilingual-v3.0
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+    - name: rerank-v3.5
+      type: reranker
+      max_input_tokens: 4096
+    - name: rerank-english-v3.0
+      type: reranker
+      max_input_tokens: 4096
+    - name: rerank-multilingual-v3.0
+      type: reranker
+      max_input_tokens: 4096
+
+- provider: xai
+  oauth:
+    client_id: b1a00492-073a-47ea-816f-4c329264a828
+    authorize_url: https://auth.x.ai/oauth2/authorize
+    token_url: https://auth.x.ai/oauth2/token
+    scopes:
+      - openid
+      - profile
+      - email
+      - offline_access
+      - grok-cli:access
+      - api:access
+    redirect_port: 56121
+    flow: pkce
+    token_request_format: form_url_encoded
+    extra_authorize_params:
+      plan: generic
+      referrer: coyote
+    echo_pkce_in_token_exchange: true
+  models:
+    - name: grok-4.20
+      real_name: grok-4.20-multi-agent-0309
+      input_price: 1.25
+      output_price: 2.5
+      max_input_tokens: 1000000
+      supports_function_calling: true
+    - name: grok-4.20-reasoning
+      real_name: grok-4.20-0309-reasoning
+      input_price: 1.25
+      output_price: 2.5
+      max_input_tokens: 1000000
+      supports_function_calling: true
+    - name: grok-4.20-non-reasoning
+      real_name: grok-4.20-0309-non-reasoning
+      input_price: 1.25
+      output_price: 2.5
+      max_input_tokens: 1000000
+      supports_function_calling: true
+
+- provider: vertexai
+  rules:
+    - match: "claude-*"
+      require_max_tokens: true
+    - match: "gemini-3.5-flash"
+      default_reasoning_effort: medium
+    - match: "gemini-3-flash-preview"
+      default_reasoning_effort: high
+    - match: "gemini-3.1-flash-lite"
+      default_reasoning_effort: minimal
+    - match: "gemini-3.1-pro-preview"
+      default_reasoning_effort: high
+    - match: "gemini-3-pro-preview"
+      default_reasoning_effort: high
+    - match: "gemini-2.5-flash"
+      default_reasoning_effort: medium
+    - match: "gemini-2.5-pro"
+      default_reasoning_effort: high
+    - match: "claude-opus-5"
+      default_reasoning_effort: high
+    - match: "claude-fable-*"
+      default_reasoning_effort: high
+    - match: "claude-opus-4-8"
+      default_reasoning_effort: high
+    - match: "claude-opus-4-7"
+      default_reasoning_effort: high
+    - match: "claude-sonnet-5"
+      default_reasoning_effort: high
+    - match: "claude-sonnet-4-6"
+      default_reasoning_effort: high
+  variants:
+    - match: "claude-opus-4-6"
+      suffix: ":thinking"
+      real_name: claude-opus-4-6
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-sonnet-4-6"
+      suffix: ":thinking"
+      real_name: claude-sonnet-4-6
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-opus-4-5@20251101"
+      suffix: ":thinking"
+      real_name: claude-opus-4-5@20251101
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-sonnet-4-5@20250929"
+      suffix: ":thinking"
+      real_name: claude-sonnet-4-5@20250929
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+    - match: "claude-haiku-4-5@20251001"
+      suffix: ":thinking"
+      real_name: claude-haiku-4-5@20251001
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          temperature: null
+          top_p: null
+          thinking:
+            type: enabled
+            budget_tokens: 16000
+  models:
+    - name: text-embedding-005
+      type: embedding
+      max_input_tokens: 20000
+      input_price: 0.025
+      max_tokens_per_chunk: 2048
+      default_chunk_size: 1500
+      max_batch_size: 5
+    - name: text-multilingual-embedding-002
+      type: embedding
+      max_input_tokens: 20000
+      input_price: 0.2
+      max_tokens_per_chunk: 2048
+      default_chunk_size: 1500
+      max_batch_size: 5
+
+- provider: bedrock
+  rules:
+    - match: "us.anthropic.claude-*"
+      require_max_tokens: true
+    - match: "us.meta.llama*"
+      require_max_tokens: true
+    - match: "us.anthropic.claude-opus-5"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-fable-*"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-opus-4-8"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-opus-4-7"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-opus-4-6-v1"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-sonnet-5"
+      default_reasoning_effort: high
+    - match: "us.anthropic.claude-sonnet-4-6"
+      default_reasoning_effort: high
+  variants:
+    - match: "us.anthropic.claude-opus-4-6-v1"
+      suffix: ":thinking"
+      real_name: us.anthropic.claude-opus-4-6-v1
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          inferenceConfig:
+            temperature: null
+            topP: null
+          additionalModelRequestFields:
+            thinking:
+              type: enabled
+              budget_tokens: 16000
+    - match: "us.anthropic.claude-sonnet-4-6"
+      suffix: ":thinking"
+      real_name: us.anthropic.claude-sonnet-4-6
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          inferenceConfig:
+            temperature: null
+            topP: null
+          additionalModelRequestFields:
+            thinking:
+              type: enabled
+              budget_tokens: 16000
+    - match: "us.anthropic.claude-opus-4-5-20251101-v1:0"
+      suffix: ":thinking"
+      real_name: us.anthropic.claude-opus-4-5-20251101-v1:0
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          inferenceConfig:
+            temperature: null
+            topP: null
+          additionalModelRequestFields:
+            thinking:
+              type: enabled
+              budget_tokens: 16000
+    - match: "us.anthropic.claude-sonnet-4-5-20250929-v1:0"
+      suffix: ":thinking"
+      real_name: us.anthropic.claude-sonnet-4-5-20250929-v1:0
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          inferenceConfig:
+            temperature: null
+            topP: null
+          additionalModelRequestFields:
+            thinking:
+              type: enabled
+              budget_tokens: 16000
+    - match: "us.anthropic.claude-haiku-4-5-20251001-v1:0"
+      suffix: ":thinking"
+      real_name: us.anthropic.claude-haiku-4-5-20251001-v1:0
+      max_output_tokens: 24000
+      supports_function_calling: false
+      reasoning_levels: []
+      patch:
+        body:
+          inferenceConfig:
+            temperature: null
+            topP: null
+          additionalModelRequestFields:
+            thinking:
+              type: enabled
+              budget_tokens: 16000
+  models:
+    - name: cohere.embed-english-v3
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+    - name: cohere.embed-multilingual-v3
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+
+- provider: cloudflare
+  rules:
+    - match: "@cf/*"
+      require_max_tokens: true
+  models:
+    - name: '@cf/baai/bge-large-en-v1.5'
+      type: embedding
+      input_price: 0
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 100
+
+- provider: ernie
+  models:
+    - name: ernie-4.5-turbo-128k
+      max_input_tokens: 131072
+      input_price: 0.112
+      output_price: 0.448
+    - name: ernie-4.5-turbo-vl-32k
+      max_input_tokens: 32768
+      input_price: 0.42
+      output_price: 1.26
+      supports_vision: true
+    - name: ernie-5.0-thinking-preview
+      max_input_tokens: 131072
+      input_price: 1.4
+      output_price: 5.6
+    - name: ernie-x1.1-preview
+      max_input_tokens: 65536
+      input_price: 0.14
+      output_price: 0.56
+    - name: bge-large-zh
+      type: embedding
+      input_price: 0.07
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 16
+    - name: bge-large-en
+      type: embedding
+      input_price: 0.07
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 16
+    - name: bce-reranker-base
+      type: reranker
+      max_input_tokens: 1024
+      input_price: 0.07
+
+- provider: qianwen
+  rules:
+    - match: "qwen3.5-plus"
+      patch:
+        body:
+          enable_thinking: false
+    - match: "qwen3.5-397b-a17b"
+      patch:
+        body:
+          enable_thinking: false
+  variants:
+    - match: "qwen3-max"
+      suffix: ":thinking"
+      real_name: qwen3-max
+      patch:
+        body:
+          enable_thinking: true
+  models:
+    - name: qwen3.5-plus:thinking
+      real_name: qwen3.5-plus
+      max_input_tokens: 262144
+      supports_function_calling: true
+    - name: qwen3.5-397b-a17b:thinking
+      real_name: qwen3.5-397b-a17b
+      max_input_tokens: 262144
+      supports_function_calling: true
+    - name: text-embedding-v4
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 2000
+      max_batch_size: 10
+    - name: text-embedding-v3
+      type: embedding
+      input_price: 0.1
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 2000
+      max_batch_size: 10
+
+- provider: hunyuan
+  models:
+    - name: hunyuan-2.0-instruct-20251111
+      max_input_tokens: 131072
+      input_price: 0.112
+      output_price: 0.28
+      supports_function_calling: true
+    - name: hunyuan-2.0-thinking-20251109
+      max_input_tokens: 131072
+      input_price: 0.14
+      output_price: 0.56
+      supports_function_calling: true
+    - name: hunyuan-vision-1.5-instruct
+      max_input_tokens: 24576
+      input_price: 0.42
+      output_price: 1.26
+      supports_vision: true
+    - name: hunyuan-embedding
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 1024
+      default_chunk_size: 1000
+      max_batch_size: 100
+
+- provider: zhipuai
+  variants:
+    - match: "glm-5"
+      suffix: ":instruct"
+      real_name: glm-5
+      patch:
+        body:
+          thinking:
+            type: disabled
+    - match: "glm-4.7"
+      suffix: ":instruct"
+      real_name: glm-4.7
+      patch:
+        body:
+          thinking:
+            type: disabled
+  models:
+    - name: embedding-3
+      type: embedding
+      max_input_tokens: 8192
+      input_price: 0.07
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 2000
+    - name: rerank
+      type: reranker
+      max_input_tokens: 4096
+      input_price: 0.112
+
+- provider: openrouter
+  rules:
+    - match: "anthropic/claude-*"
+      require_max_tokens: true
+    - match: "openai/gpt-5.6-*"
+      default_reasoning_effort: medium
+    - match: "openai/gpt-5.5"
+      default_reasoning_effort: medium
+    - match: "openai/gpt-5.5-pro"
+      default_reasoning_effort: high
+    - match: "openai/gpt-5.4"
+      default_reasoning_effort: none
+    - match: "openai/gpt-5.4-pro"
+      default_reasoning_effort: medium
+    - match: "openai/gpt-5.4-mini"
+      default_reasoning_effort: none
+    - match: "openai/gpt-5.4-nano"
+      default_reasoning_effort: none
+    - match: "openai/gpt-5.3-codex"
+      default_reasoning_effort: medium
+    - match: "openai/gpt-5.2"
+      default_reasoning_effort: none
+    - match: "openai/gpt-5.2-pro"
+      default_reasoning_effort: medium
+    - match: "openai/gpt-5"
+      default_reasoning_effort: medium
+    - match: "google/gemini-3.5-flash"
+      default_reasoning_effort: medium
+    - match: "google/gemini-3-flash-preview"
+      default_reasoning_effort: high
+    - match: "google/gemini-3.1-flash-lite"
+      default_reasoning_effort: minimal
+    - match: "google/gemini-3.1-pro-preview"
+      default_reasoning_effort: high
+    - match: "google/gemini-3-pro-preview"
+      default_reasoning_effort: high
+    - match: "google/gemini-2.5-flash"
+      default_reasoning_effort: low
+    - match: "google/gemini-2.5-pro"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-opus-5"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-fable-*"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-opus-4-8"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-opus-4-7"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-opus-4.6"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-sonnet-5"
+      default_reasoning_effort: high
+    - match: "anthropic/claude-sonnet-4.6"
+      default_reasoning_effort: high
+    - match: "perplexity/sonar-reasoning-pro"
+      patch:
+        body:
+          include_reasoning: true
+    - match: "perplexity/sonar-deep-research"
+      patch:
+        body:
+          include_reasoning: true
+
+- provider: github
+  models:
+    - name: text-embedding-3-large
+      type: embedding
+      max_tokens_per_chunk: 8191
+      default_chunk_size: 2000
+      max_batch_size: 100
+    - name: text-embedding-3-small
+      type: embedding
+      max_tokens_per_chunk: 8191
+      default_chunk_size: 2000
+      max_batch_size: 100
+    - name: cohere-embed-v3-english
+      type: embedding
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+    - name: cohere-embed-v3-multilingual
+      type: embedding
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 96
+
+- provider: deepinfra
+  models:
+    - name: BAAI/bge-large-en-v1.5
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 100
+    - name: BAAI/bge-m3
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 2000
+      max_batch_size: 100
+    - name: intfloat/e5-large-v2
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 100
+    - name: intfloat/multilingual-e5-large
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 100
+    - name: thenlper/gte-large
+      type: embedding
+      input_price: 0.01
+      max_tokens_per_chunk: 512
+      default_chunk_size: 1000
+      max_batch_size: 100
+
+- provider: jina
+  models:
+    - name: jina-embeddings-v3
+      type: embedding
+      input_price: 0
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 2000
+      max_batch_size: 100
+    - name: jina-clip-v2
+      type: embedding
+      input_price: 0
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 1500
+      max_batch_size: 100
+    - name: jina-colbert-v2
+      type: embedding
+      input_price: 0
+      max_tokens_per_chunk: 8192
+      default_chunk_size: 1500
+      max_batch_size: 100
+    - name: jina-reranker-v2-base-multilingual
+      type: reranker
+      max_input_tokens: 8192
+      input_price: 0
+    - name: jina-colbert-v2
+      type: reranker
+      max_input_tokens: 8192
+      input_price: 0
+
+- provider: voyageai
+  models:
+    - name: voyage-3-large
+      type: embedding
+      max_input_tokens: 120000
+      input_price: 0.18
+      max_tokens_per_chunk: 32000
+      default_chunk_size: 2000
+      max_batch_size: 128
+    - name: voyage-3
+      type: embedding
+      max_input_tokens: 320000
+      input_price: 0.06
+      max_tokens_per_chunk: 32000
+      default_chunk_size: 2000
+      max_batch_size: 128
+    - name: voyage-3-lite
+      type: embedding
+      max_input_tokens: 1000000
+      input_price: 0.02
+      max_tokens_per_chunk: 32000
+      default_chunk_size: 1000
+      max_batch_size: 128
+    - name: rerank-2
+      type: reranker
+      max_input_tokens: 16000
+      input_price: 0.05
+    - name: rerank-2-lite
+      type: reranker
+      max_input_tokens: 8000
+      input_price: 0.02