OpenRouter
Run any OpenRouter model by pasting its model slug.
v3
Run any vision-language model available on OpenRouter by pasting its model slug into the model_id field - e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, google/gemini-2.5-pro, qwen/qwen3.6-27b.
This is the generic escape hatch for OpenRouter - when you want a model that doesn't have a dedicated block (Qwen-VL, Kimi, Gemma, Llama Vision) and you want to try it out without waiting for a new block to be added.
The block supports the standard VLM task-type surface:
Open Prompt (
unconstrained) - Use any prompt to generate a raw responseText Recognition (OCR) (
ocr) - Model recognizes text in the imageVisual Question Answering (
visual-question-answering) - Model answers the question you submit in the promptCaptioning (short) (
caption) - Model provides a short description of the imageCaptioning (
detailed-caption) - Model provides a long description of the imageSingle-Label Classification (
classification) - Model classifies the image content as one of the provided classesMulti-Label Classification (
multi-label-classification) - Model classifies the image content as one or more of the provided classesUnprompted Object Detection (
object-detection) - Model detects and returns the bounding boxes for prominent objects in the imageStructured Output Generation (
structured-answering) - Model returns a JSON response with the specified fields
🛠️ API key
By default the block uses the Roboflow-managed OpenRouter key and bills your Roboflow credits - no extra setup needed. To bypass Roboflow billing, paste your own sk-or-... key into the api_key field.
🔒 Privacy filter
No data collection (default) – providers may not train on your inputs.
Allow data collection – broader provider pool.
Zero data retention – strictest, restricts to providers that retain nothing.
Version Differences
This version (v3) decodes model answers inside the block:
predictions- classification and object-detection answers are parsed here, so the deprecatedVLM as Detector/VLM as Classifierblocks are no longer needed. The output kind followstask_type: object detection predictions forobject-detection, classification predictions forclassificationandmulti-label-classification, andNonefor every other task.error_status-Truewhen the model answer could not be parsed.inference_id- identifier generated per image and attached to the decoded predictions.detection_format- forobject-detection, the bounding-box format the model is asked for and decoded with (xyxy/yxyx/ named keys, in 0-1, 0-1000, percent or pixels). Pick the format the model was trained on so a new model can be used without a block change; the default is the generic normalized-floats JSON.
Type identifier
Use the following identifier in step "type" field: roboflow_core/openrouter@v3 to add the block as a step in your workflow.
Properties
Name
Type
Description
Refs
name
str
Enter a unique identifier for this step..
❌
api_key
str
OpenRouter API key. Defaults to Roboflow's managed key, billed in credits via Roboflow. Provide your own sk-or-... key to call OpenRouter directly without Roboflow billing..
✅
privacy_level
str
Provider privacy filter. Stricter levels reduce the pool of providers and may increase per-call cost on the managed key..
❌
max_tokens
int
Maximum number of tokens the model can generate in its response. Defaults to 2048. Raise it explicitly (e.g. 8192) when a task needs a longer answer. Billing is based on tokens actually generated, not on this limit..
❌
temperature
float
Temperature to sample from the model - value in range 0.0-2.0, the higher - the more random / "creative" the generations are..
✅
max_concurrent_requests
int
Number of concurrent requests for batches of images. If not given - block defaults to value configured globally in Workflows Execution Engine. Restrict if you hit rate limits..
❌
model_id
str
OpenRouter model slug, e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, qwen/qwen3.6-27b. See https://openrouter.ai/models for the full list..
✅
reasoning_effort
str
Extended-reasoning budget forwarded to OpenRouter as reasoning: {"effort": ...}. Unset keeps the model's provider-default behavior; none explicitly disables reasoning. Models that reject the config are retried without it. Reasoning tokens count toward max_tokens, so raise it for medium/high..
✅
task_type
str
Task type to be performed by model. Value determines required parameters and output response..
❌
prompt
str
Text prompt to send to the model..
✅
output_structure
Dict[str, str]
Dictionary with structure of expected JSON response..
❌
classes
List[str]
List of classes to be used..
✅
detection_format
str
Bounding-box format for object-detection: how the model is asked for boxes and how its answer is decoded. Pick the format the model was trained on; keep the default when unsure..
✅
The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.
Runtime compatibility
requires_internet - air-gapped / offline deployments : This block depends on a service that is not reachable from fully offline / air-gapped deployments.
Input and Output Bindings
The available connections depend on its binding kinds. Check what binding kinds OpenRouter in version v3 has.
v2
Run any vision-language model available on OpenRouter by pasting its model slug into the model_id field - e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, google/gemini-2.5-pro, qwen/qwen3.6-27b.
This is the generic escape hatch for OpenRouter - when you want a model that doesn't have a dedicated block (Qwen-VL, Kimi, Gemma, Llama Vision) and you want to try it out without waiting for a new block to be added.
The block supports the standard VLM task-type surface:
Open Prompt (
unconstrained) - Use any prompt to generate a raw responseText Recognition (OCR) (
ocr) - Model recognizes text in the imageVisual Question Answering (
visual-question-answering) - Model answers the question you submit in the promptCaptioning (short) (
caption) - Model provides a short description of the imageCaptioning (
detailed-caption) - Model provides a long description of the imageSingle-Label Classification (
classification) - Model classifies the image content as one of the provided classesMulti-Label Classification (
multi-label-classification) - Model classifies the image content as one or more of the provided classesUnprompted Object Detection (
object-detection) - Model detects and returns the bounding boxes for prominent objects in the imageStructured Output Generation (
structured-answering) - Model returns a JSON response with the specified fields
🛠️ API key
By default the block uses the Roboflow-managed OpenRouter key and bills your Roboflow credits - no extra setup needed. To bypass Roboflow billing, paste your own sk-or-... key into the api_key field.
🔒 Privacy filter
No data collection (default) – providers may not train on your inputs.
Allow data collection – broader provider pool.
Zero data retention – strictest, restricts to providers that retain nothing.
Type identifier
Use the following identifier in step "type" field: roboflow_core/openrouter@v2 to add the block as a step in your workflow.
Properties
Name
Type
Description
Refs
name
str
Enter a unique identifier for this step..
❌
api_key
str
OpenRouter API key. Defaults to Roboflow's managed key, billed in credits via Roboflow. Provide your own sk-or-... key to call OpenRouter directly without Roboflow billing..
✅
privacy_level
str
Provider privacy filter. Stricter levels reduce the pool of providers and may increase per-call cost on the managed key..
❌
max_tokens
int
Maximum number of tokens the model can generate in its response. Defaults to 2048. Raise it explicitly (e.g. 8192) when a task needs a longer answer. Billing is based on tokens actually generated, not on this limit..
❌
temperature
float
Temperature to sample from the model - value in range 0.0-2.0, the higher - the more random / "creative" the generations are..
✅
max_concurrent_requests
int
Number of concurrent requests for batches of images. If not given - block defaults to value configured globally in Workflows Execution Engine. Restrict if you hit rate limits..
❌
model_id
str
OpenRouter model slug, e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, qwen/qwen3.6-27b. See https://openrouter.ai/models for the full list..
✅
reasoning_effort
str
Extended-reasoning budget forwarded to OpenRouter as reasoning: {"effort": ...}. Unset keeps the model's provider-default behavior; none explicitly disables reasoning. Models that reject the config are retried without it. Reasoning tokens count toward max_tokens, so raise it for medium/high..
✅
task_type
str
Task type to be performed by model. Value determines required parameters and output response..
❌
prompt
str
Text prompt to send to the model..
✅
output_structure
Dict[str, str]
Dictionary with structure of expected JSON response..
❌
classes
List[str]
List of classes to be used..
✅
The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.
Runtime compatibility
requires_internet - air-gapped / offline deployments : This block depends on a service that is not reachable from fully offline / air-gapped deployments.
Input and Output Bindings
The available connections depend on its binding kinds. Check what binding kinds OpenRouter in version v2 has.
v1
Run any vision-language model available on OpenRouter by pasting its model slug into the model_id field - e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, google/gemini-2.5-pro, qwen/qwen3.6-27b.
This is the generic escape hatch for OpenRouter - when you want a model that doesn't have a dedicated block (Qwen-VL, Kimi, Gemma, Llama Vision) and you want to try it out without waiting for a new block to be added.
The block supports the standard VLM task-type surface:
Open Prompt (
unconstrained) - Use any prompt to generate a raw responseText Recognition (OCR) (
ocr) - Model recognizes text in the imageVisual Question Answering (
visual-question-answering) - Model answers the question you submit in the promptCaptioning (short) (
caption) - Model provides a short description of the imageCaptioning (
detailed-caption) - Model provides a long description of the imageSingle-Label Classification (
classification) - Model classifies the image content as one of the provided classesMulti-Label Classification (
multi-label-classification) - Model classifies the image content as one or more of the provided classesUnprompted Object Detection (
object-detection) - Model detects and returns the bounding boxes for prominent objects in the imageStructured Output Generation (
structured-answering) - Model returns a JSON response with the specified fields
🛠️ API key
By default the block uses the Roboflow-managed OpenRouter key and bills your Roboflow credits - no extra setup needed. To bypass Roboflow billing, paste your own sk-or-... key into the api_key field.
🔒 Privacy filter
No data collection (default) – providers may not train on your inputs.
Allow data collection – broader provider pool.
Zero data retention – strictest, restricts to providers that retain nothing.
Type identifier
Use the following identifier in step "type" field: roboflow_core/openrouter@v1 to add the block as a step in your workflow.
Properties
Name
Type
Description
Refs
name
str
Enter a unique identifier for this step..
❌
api_key
str
OpenRouter API key. Defaults to Roboflow's managed key, billed in credits via Roboflow. Provide your own sk-or-... key to call OpenRouter directly without Roboflow billing..
✅
privacy_level
str
Provider privacy filter. Stricter levels reduce the pool of providers and may increase per-call cost on the managed key..
❌
max_tokens
int
Maximum number of tokens the model can generate in its response..
❌
temperature
float
Temperature to sample from the model - value in range 0.0-2.0, the higher - the more random / "creative" the generations are..
✅
max_concurrent_requests
int
Number of concurrent requests for batches of images. If not given - block defaults to value configured globally in Workflows Execution Engine. Restrict if you hit rate limits..
❌
model_id
str
OpenRouter model slug, e.g. openai/gpt-4o-mini, anthropic/claude-3.5-sonnet, qwen/qwen3.6-27b. See https://openrouter.ai/models for the full list..
✅
task_type
str
Task type to be performed by model. Value determines required parameters and output response..
❌
prompt
str
Text prompt to send to the model..
✅
output_structure
Dict[str, str]
Dictionary with structure of expected JSON response..
❌
classes
List[str]
List of classes to be used..
✅
The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.
Runtime compatibility
requires_internet - air-gapped / offline deployments : This block depends on a service that is not reachable from fully offline / air-gapped deployments.
Input and Output Bindings
The available connections depend on its binding kinds. Check what binding kinds OpenRouter in version v1 has.
Last updated
Was this helpful?