Qwen2-VL-2B-Instruct is Qwen’s instruction-tuned 2B vision-language model for image-and-text questions. The producer describes handling images of different resolutions and aspect ratios and understanding text in several languages inside pictures. Its generated result is a written response to the supplied visual material and question. The manufacturer also describes visual-agent use with devices such as phones and robots. For that workflow, supply the current visual environment and a text instruction; the surrounding application carries out the proposed operation and returns the observed state for the next question.
Choose the image containing the evidence needed for the answer, then identify the detail to interpret. A document question can ask about a visible passage; a chart question can identify the series, axes and comparison to explain. Keep the image identifier with the requested result so the response remains tied to the correct reference.
The maker lists image-text understanding beyond English and Chinese, including European languages, Japanese, Korean, Arabic and Vietnamese. Preserve the original wording and script when asking for a transcription, explanation or translation of visible text. An illustrative brief can request the displayed label and a separate explanation of its meaning.
Qwen2-VL uses dynamic resolution to map an image to a variable number of visual tokens. In the native processor example, the default range is 4 to 16,384 visual tokens per image. Its native guide illustrates minimum and maximum pixel settings, including a narrower image-token preparation range of 256 to 1,280, to balance detail with processing needs. Those settings retain the aspect ratio when resizing.
The producer also permits requested resized height and width, rounded to a multiple of 28. Select a preparation that leaves the important labels readable, and retain the original image when using a crop. The native examples include multiple images and batches, allowing a question to keep its separate references identified.
The maker identifies limitations in complex multi-step instructions, object counting in complicated scenes and spatial reasoning in three dimensions. It also notes incomplete recognition of particular people or brands. Ask for a focused visual observation and inspect the particular count, label or relative position the result depends on.
The producer states that the image dataset was updated through June 2023. Provide the written context needed for a recent or unfamiliar label rather than asking the picture alone to establish that information. Keep a quoted label distinguishable from an inferred identity in the returned note.
The published checkpoint configuration sets 32,768 maximum positions. Prepare the question and visual inputs with the matching model processor, retaining the image order used for the request. The native helper accepts local images, base64 and image addresses.
Save the selected 2B Instruct name, references and completed written answer together. For a follow-up, include the relevant material and identify the particular observation or interpretation to revise, then compare the corrected result with that source.
Through Ailoot, use the exact model ID in /v1/chat/completions and keep the image with its question in the conversation. The image content form uses image_url; supported image sources can carry an image address or base64 data URI. Collect the full JSON answer or all SSE segments when stream is enabled, and retain usage with the completed result.
", "aboutText": "Qwen2-VL-2B-Instruct is Qwen’s instruction-tuned 2B vision-language model for image-and-text questions. The producer describes handling images of different resolutions and aspect ratios and understanding text in several languages inside pictures. Its generated result is a written response to the supplied visual material and question. The manufacturer also describes visual-agent use with devices such as phones and robots. For that workflow, supply the current visual environment and a text instruction; the surrounding application carries out the proposed operation and returns the observed state for the next question.\n\n### Supply the picture and a precise question\n\nChoose the image containing the evidence needed for the answer, then identify the detail to interpret. A document question can ask about a visible passage; a chart question can identify the series, axes and comparison to explain. Keep the image identifier with the requested result so the response remains tied to the correct reference.\n\nThe maker lists image-text understanding beyond English and Chinese, including European languages, Japanese, Korean, Arabic and Vietnamese. Preserve the original wording and script when asking for a transcription, explanation or translation of visible text. An illustrative brief can request the displayed label and a separate explanation of its meaning.\n\n### Prepare resolution and visual tokens deliberately\n\nQwen2-VL uses dynamic resolution to map an image to a variable number of visual tokens. In the native processor example, the default range is 4 to 16,384 visual tokens per image. Its native guide illustrates minimum and maximum pixel settings, including a narrower image-token preparation range of 256 to 1,280, to balance detail with processing needs. Those settings retain the aspect ratio when resizing.\n\nThe producer also permits requested resized height and width, rounded to a multiple of 28. Select a preparation that leaves the important labels readable, and retain the original image when using a crop. The native examples include multiple images and batches, allowing a question to keep its separate references identified.\n\n### Work within the documented visual limitations\n\nThe maker identifies limitations in complex multi-step instructions, object counting in complicated scenes and spatial reasoning in three dimensions. It also notes incomplete recognition of particular people or brands. Ask for a focused visual observation and inspect the particular count, label or relative position the result depends on.\n\nThe producer states that the image dataset was updated through June 2023. Provide the written context needed for a recent or unfamiliar label rather than asking the picture alone to establish that information. Keep a quoted label distinguishable from an inferred identity in the returned note.\n\n### Keep the prepared inputs with the answer\n\nThe published checkpoint configuration sets 32,768 maximum positions. Prepare the question and visual inputs with the matching model processor, retaining the image order used for the request. The native helper accepts local images, base64 and image addresses.\n\nSave the selected 2B Instruct name, references and completed written answer together. For a follow-up, include the relevant material and identify the particular observation or interpretation to revise, then compare the corrected result with that source.\n\nThrough Ailoot, use the exact model ID in /v1/chat/completions and keep the image with its question in the conversation. The image content form uses image_url; supported image sources can carry an image address or base64 data URI. Collect the full JSON answer or all SSE segments when stream is enabled, and retain usage with the completed result.", "editorialOverlay": { "copy_fields": null, "copy_sha256": null, "copy_source": null, "html_sha256": "23abef244c9b79a98cef15102c3cd16fea06fd4b8e36bf7801102c7bb6ed4ab0", "module_sha256": "ee9dc1c50b1e6a8fb7fa062030fcb4646b9aaaeb08be513b723a5a92a7290f06", "raw_model_unchanged": true, "schema": "ailoot-native-model-editorial-v1" }, "locale": "en", "methods": [ { "html": "qwen2-vl-2b-instructQwen2-VL-2B-Instruct is Qwen’s vision-language model for image questions, multilingual text in pictures and visual explanations. Supply a readable image and a precise question for a written result tied to the visible evidence.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen2-VL-2B-Instruct is Qwen’s instruction-tuned 2B vision-language model for image-and-text questions. The producer describes handling images of different resolutions and aspect ratios and understanding text in several languages inside pictures. Its generated result is a written response to the supplied visual material and question. The manufacturer also describes visual-agent use with devices such as phones and robots. For that workflow, supply the current visual environment and a text instruction; the surrounding application carries out the proposed operation and returns the observed state for the next question.
Choose the image containing the evidence needed for the answer, then identify the detail to interpret. A document question can ask about a visible passage; a chart question can identify the series, axes and comparison to explain. Keep the image identifier with the requested result so the response remains tied to the correct reference.
The maker lists image-text understanding beyond English and Chinese, including European languages, Japanese, Korean, Arabic and Vietnamese. Preserve the original wording and script when asking for a transcription, explanation or translation of visible text. An illustrative brief can request the displayed label and a separate explanation of its meaning.
Qwen2-VL uses dynamic resolution to map an image to a variable number of visual tokens. In the native processor example, the default range is 4 to 16,384 visual tokens per image. Its native guide illustrates minimum and maximum pixel settings, including a narrower image-token preparation range of 256 to 1,280, to balance detail with processing needs. Those settings retain the aspect ratio when resizing.
The producer also permits requested resized height and width, rounded to a multiple of 28. Select a preparation that leaves the important labels readable, and retain the original image when using a crop. The native examples include multiple images and batches, allowing a question to keep its separate references identified.
The maker identifies limitations in complex multi-step instructions, object counting in complicated scenes and spatial reasoning in three dimensions. It also notes incomplete recognition of particular people or brands. Ask for a focused visual observation and inspect the particular count, label or relative position the result depends on.
The producer states that the image dataset was updated through June 2023. Provide the written context needed for a recent or unfamiliar label rather than asking the picture alone to establish that information. Keep a quoted label distinguishable from an inferred identity in the returned note.
The published checkpoint configuration sets 32,768 maximum positions. Prepare the question and visual inputs with the matching model processor, retaining the image order used for the request. The native helper accepts local images, base64 and image addresses.
Save the selected 2B Instruct name, references and completed written answer together. For a follow-up, include the relevant material and identify the particular observation or interpretation to revise, then compare the corrected result with that source.
Through Ailoot, use the exact model ID in /v1/chat/completions and keep the image with its question in the conversation. The image content form uses image_url; supported image sources can carry an image address or base64 data URI. Collect the full JSON answer or all SSE segments when stream is enabled, and retain usage with the completed result.
POST /v1/chat/completionstext, imagetext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen2-vl-2b-instruct Conditions: Use this exact Ailoot routing selector. | Default— |
Fieldmessages | Typearray | RequirementRequired | Values & limits Conditions: | Default— |
Fieldstream | Typeboolean | RequirementOptional | Values & limits Conditions: | Defaultfalse |
text_input · Required: messages
image_with_text · Required: messages with actual image content blocks
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, messages |
Parameteroptional | Contractstream |
Parametermaterial_types | Contracttext, image |
Parameterconditions | ContractUse this exact Ailoot routing selector., Image content is carried by this protocol where the exact selected model supports vision. Protocol admission does not establish image capability for every model. |
Method aliases: /chat/completions
Lifecycle: synchronous_or_streamed
Tariff references do not add request options. Supported parameters are listed for the selected method.
| Configuration | Rate schedule | Charge |
|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Charge Input · Input tokens $0 / 1K tokens Output · Output tokens $0 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\nexport AILOOT_API_KEY\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "qwen2-vl-2b-instruct", "messages": [{"role": "user", "content": "Explain how a rainbow forms."}]}')\n\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/chat/completions \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\n# Submit once. Retain the complete result, original identifiers and every response header.\nJSON or SSE
Keep every complete returned response, file and content part.
Retain every content block, tool-call ID, thinking signature when returned, all SSE deltas and complete usage. A first text fragment is not the whole result.
Keep the original X-Client-Request-ID, response and headers. An unknown or result_pending_check outcome does not authorize a second creation.
| Field | Type |
|---|---|
Fieldchoices[] | Typeobject |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionA required field, model selector, type or input representation is invalid. | Next stepInspect the returned validation detail before preparing a separate request. |
Response | ConditionThe original request or result is unknown or requires a check. | Next stepKeep original identifiers and headers; read only supported status/library records. Do not submit creation again. |
Ailoot Ultra qwen2-vl-2b-instruct
/chat/completions
qwen2-vl-2b-instructQwen2-VL-2B-Instruct is Qwen’s vision-language model for image questions, multilingual text in pictures and visual explanations. Supply a readable image and a precise question for a written result tied to the visible evidence.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen2-VL-2B-Instruct is Qwen’s instruction-tuned 2B vision-language model for image-and-text questions. The producer describes handling images of different resolutions and aspect ratios and understanding text in several languages inside pictures. Its generated result is a written response to the supplied visual material and question. The manufacturer also describes visual-agent use with devices such as phones and robots. For that workflow, supply the current visual environment and a text instruction; the surrounding application carries out the proposed operation and returns the observed state for the next question.
Choose the image containing the evidence needed for the answer, then identify the detail to interpret. A document question can ask about a visible passage; a chart question can identify the series, axes and comparison to explain. Keep the image identifier with the requested result so the response remains tied to the correct reference.
The maker lists image-text understanding beyond English and Chinese, including European languages, Japanese, Korean, Arabic and Vietnamese. Preserve the original wording and script when asking for a transcription, explanation or translation of visible text. An illustrative brief can request the displayed label and a separate explanation of its meaning.
Qwen2-VL uses dynamic resolution to map an image to a variable number of visual tokens. In the native processor example, the default range is 4 to 16,384 visual tokens per image. Its native guide illustrates minimum and maximum pixel settings, including a narrower image-token preparation range of 256 to 1,280, to balance detail with processing needs. Those settings retain the aspect ratio when resizing.
The producer also permits requested resized height and width, rounded to a multiple of 28. Select a preparation that leaves the important labels readable, and retain the original image when using a crop. The native examples include multiple images and batches, allowing a question to keep its separate references identified.
The maker identifies limitations in complex multi-step instructions, object counting in complicated scenes and spatial reasoning in three dimensions. It also notes incomplete recognition of particular people or brands. Ask for a focused visual observation and inspect the particular count, label or relative position the result depends on.
The producer states that the image dataset was updated through June 2023. Provide the written context needed for a recent or unfamiliar label rather than asking the picture alone to establish that information. Keep a quoted label distinguishable from an inferred identity in the returned note.
The published checkpoint configuration sets 32,768 maximum positions. Prepare the question and visual inputs with the matching model processor, retaining the image order used for the request. The native helper accepts local images, base64 and image addresses.
Save the selected 2B Instruct name, references and completed written answer together. For a follow-up, include the relevant material and identify the particular observation or interpretation to revise, then compare the corrected result with that source.
Through Ailoot, use the exact model ID in /v1/chat/completions and keep the image with its question in the conversation. The image content form uses image_url; supported image sources can carry an image address or base64 data URI. Collect the full JSON answer or all SSE segments when stream is enabled, and retain usage with the completed result.
POST /v1/messagestext, imagejson, sse| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen2-vl-2b-instruct Conditions: Use this exact Ailoot routing selector. | Default— |
Fieldmessages | Typearray | RequirementRequired | Values & limits Conditions: | Default— |
Fieldstream | Typeboolean | RequirementOptional | Values & limits Conditions: | Defaultfalse |
Fieldmax_tokens | Typeinteger | RequirementRequired | Values & limits Minimum: 1 Conditions: | Default— |
text_input · Required: messages
image_with_text · Required: messages with actual image content blocks
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, messages, max_tokens |
Parameteroptional | Contractstream |
Parametermaterial_types | Contracttext, image |
Parameterconditions | ContractUse this exact Ailoot routing selector., Image content is carried by this protocol where the exact selected model supports vision. Protocol admission does not establish image capability for every model. |
Lifecycle: synchronous_or_streamed
Tariff references do not add request options. Supported parameters are listed for the selected method.
| Configuration | Rate schedule | Charge |
|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Charge Input · Input tokens $0 / 1K tokens Output · Output tokens $0 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\nexport AILOOT_API_KEY\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "qwen2-vl-2b-instruct", "messages": [{"role": "user", "content": "Explain how a rainbow forms."}], "max_tokens": 256}')\n\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/messages \\\n -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\n# Submit once. Retain the complete result, original identifiers and every response header.\nJSON or SSE
Keep every complete returned response, file and content part.
Retain every content block, tool-call ID, thinking signature when returned, all SSE deltas and complete usage. A first text fragment is not the whole result.
Keep the original X-Client-Request-ID, response and headers. An unknown or result_pending_check outcome does not authorize a second creation.
| Field | Type |
|---|---|
Fieldcontent[] | Typeobject |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionA required field, model selector, type or input representation is invalid. | Next stepInspect the returned validation detail before preparing a separate request. |
Response | ConditionThe original request or result is unknown or requires a check. | Next stepKeep original identifiers and headers; read only supported status/library records. Do not submit creation again. |
Ailoot Ultra qwen2-vl-2b-instruct
/chat/completions