Introduction: the instruction-tuned 3B edition
Qwen2.5-VL-3B-Instruct is the 3B member of Qwen2.5-VL. It accepts text instructions with images or video and generates textual responses, including structured data. Choose it for a clearly defined visual question, a document-extraction task or an explanation that connects visible content with the user’s request. This exact size and release retain their own identity; features added specifically to other sizes or later Qwen generations should not be assumed to apply to this checkpoint.
Visual understanding and agent workflows
Qwen describes recognition of common objects and analysis of writing, charts, icons, graphics and page layouts. For a form, specify the required fields and the meaning of an empty value. For a chart, ask for the axes, units and visible values supporting the conclusion. A focused request makes the output easier to compare with the original image than a broad request to describe everything.
The family also supports visual-agent reasoning and tool direction for computer or phone workflows. The model can interpret an interface and help decide an action; connected application tools perform that action. A screenshot review can ask which control corresponds to a task and what visible information supports the choice. That use of visual reasoning does not itself grant control of a device.
Long video and event localization
The maker describes understanding videos longer than one hour and identifying the segments relevant to a requested event. This is a family capability, not an unconditional media-upload limit or a guarantee that every frame is retained. Prepare the relevant clip, preserve its timing information and ask a precise question about the event, sequence or change. The model’s video-processing approach uses frame-rate information to align visual observations with absolute time.
Localization and structured documents
Qwen2.5-VL can return object locations as bounding boxes or points, with coordinates and attributes in JSON. For a localization task, name the target object and specify whether the result should be a point or a box. Keep the coordinate convention explicit in the consuming application rather than infer a physical measurement from an arbitrary photograph.
The maker also documents structured extraction from scanned invoices, forms and tables. Provide a readable page and an output schema: for example, field name, extracted text and the visible region that supports it. Preserve units, decimal separators and table headings in the source. Treat uncertain or illegible content as a review case rather than asking the model to invent a missing value. These preparation examples apply documented capabilities; they are not guarantees of exact extraction.
Architecture changes
Dynamic image resolution is extended into video through variable frame-rate training. Updated multimodal positional encoding represents temporal sequence and absolute time, helping identify specific moments and changes in speed. The vision encoder uses window attention together with SwiGLU and RMSNorm. These changes describe how the model processes visual material; they do not establish a fixed latency for a particular application.
Preparing images and video
The official examples cover single images, multiple images, text-only conversations, video files and ordered frame lists. Label several visual inputs so a question can refer to each one. For an image comparison, request a table organized by the labels and restrict the result to visible differences. For a video timeline, preserve frame order and timing instead of presenting disconnected pictures as if they carried a known duration.
The image processor’s documented default ranges from four to 16,384 visual tokens per image. Its example can target 256–1,280 tokens by setting pixel bounds. It maintains aspect ratio within those bounds, while explicit resized dimensions are aligned to multiples of 28. More detail consumes more computation and sequence space, so a small text field may need a higher-resolution crop while a broad scene summary may not. Those are maker processing controls; application upload options can differ.
Context and substantive limits
The published checkpoint configuration sets maximum positions to 128,000, while the accompanying long-text section still discusses a 32,768-token starting configuration. The guide describes YaRN extrapolation but warns that it can significantly affect temporal and spatial localization and does not recommend that method for those tasks. It separately notes that video positional encoding can allow a larger position setting, giving 64K as an example. These different configuration descriptions should not be collapsed into a single universally validated input ceiling.
Text, visual material and the answer share a finite sequence budget. Keep relevant content legible, reserve room for the result and check extracted text, coordinates and event times against the original. The card’s 128-token generation example is illustrative and does not establish a separate hard output maximum. This model returns textual analysis or structured information; its vision-language capabilities do not mean it generates images or video.
", "aboutText": "Introduction: the instruction-tuned 3B edition\n\nQwen2.5-VL-3B-Instruct is the 3B member of Qwen2.5-VL. It accepts text instructions with images or video and generates textual responses, including structured data. Choose it for a clearly defined visual question, a document-extraction task or an explanation that connects visible content with the user’s request. This exact size and release retain their own identity; features added specifically to other sizes or later Qwen generations should not be assumed to apply to this checkpoint.\n\nVisual understanding and agent workflows\n\nQwen describes recognition of common objects and analysis of writing, charts, icons, graphics and page layouts. For a form, specify the required fields and the meaning of an empty value. For a chart, ask for the axes, units and visible values supporting the conclusion. A focused request makes the output easier to compare with the original image than a broad request to describe everything.\n\nThe family also supports visual-agent reasoning and tool direction for computer or phone workflows. The model can interpret an interface and help decide an action; connected application tools perform that action. A screenshot review can ask which control corresponds to a task and what visible information supports the choice. That use of visual reasoning does not itself grant control of a device.\n\nLong video and event localization\n\nThe maker describes understanding videos longer than one hour and identifying the segments relevant to a requested event. This is a family capability, not an unconditional media-upload limit or a guarantee that every frame is retained. Prepare the relevant clip, preserve its timing information and ask a precise question about the event, sequence or change. The model’s video-processing approach uses frame-rate information to align visual observations with absolute time.\n\nLocalization and structured documents\n\nQwen2.5-VL can return object locations as bounding boxes or points, with coordinates and attributes in JSON. For a localization task, name the target object and specify whether the result should be a point or a box. Keep the coordinate convention explicit in the consuming application rather than infer a physical measurement from an arbitrary photograph.\n\nThe maker also documents structured extraction from scanned invoices, forms and tables. Provide a readable page and an output schema: for example, field name, extracted text and the visible region that supports it. Preserve units, decimal separators and table headings in the source. Treat uncertain or illegible content as a review case rather than asking the model to invent a missing value. These preparation examples apply documented capabilities; they are not guarantees of exact extraction.\n\nArchitecture changes\n\nDynamic image resolution is extended into video through variable frame-rate training. Updated multimodal positional encoding represents temporal sequence and absolute time, helping identify specific moments and changes in speed. The vision encoder uses window attention together with SwiGLU and RMSNorm. These changes describe how the model processes visual material; they do not establish a fixed latency for a particular application.\n\nPreparing images and video\n\nThe official examples cover single images, multiple images, text-only conversations, video files and ordered frame lists. Label several visual inputs so a question can refer to each one. For an image comparison, request a table organized by the labels and restrict the result to visible differences. For a video timeline, preserve frame order and timing instead of presenting disconnected pictures as if they carried a known duration.\n\nThe image processor’s documented default ranges from four to 16,384 visual tokens per image. Its example can target 256–1,280 tokens by setting pixel bounds. It maintains aspect ratio within those bounds, while explicit resized dimensions are aligned to multiples of 28. More detail consumes more computation and sequence space, so a small text field may need a higher-resolution crop while a broad scene summary may not. Those are maker processing controls; application upload options can differ.\n\nContext and substantive limits\n\nThe published checkpoint configuration sets maximum positions to 128,000, while the accompanying long-text section still discusses a 32,768-token starting configuration. The guide describes YaRN extrapolation but warns that it can significantly affect temporal and spatial localization and does not recommend that method for those tasks. It separately notes that video positional encoding can allow a larger position setting, giving 64K as an example. These different configuration descriptions should not be collapsed into a single universally validated input ceiling.\n\nText, visual material and the answer share a finite sequence budget. Keep relevant content legible, reserve room for the result and check extracted text, coordinates and event times against the original. The card’s 128-token generation example is illustrative and does not establish a separate hard output maximum. This model returns textual analysis or structured information; its vision-language capabilities do not mean it generates images or video.", "editorialOverlay": { "copy_fields": { "practical_html": "Technical example in English
Organize a recorded event timeline\n\nUse only these text annotations from a clip: 00:02 worker picks up a box; 00:05 worker places the box on the scale; 00:08 scale display is obscured; 00:11 worker removes the box. Return the event sequence, the first annotated placement time and whether the weight is established. Preserve timestamps without claiming frame-level inspection.An ordered event table, a placement-time field and an unresolved weight field.Check that obscured weight remains unknown and the placement timestamp is attributed to the supplied annotation.qwen2.5-vl-3b-instructQwen2.5-VL-3B-Instruct is Qwen’s compact instruction-tuned vision-language model for text, images and video. It returns text and structured visual information, including document fields, object coordinates and event descriptions, with dynamic resolution and time-aware video processing.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Introduction: the instruction-tuned 3B edition
Qwen2.5-VL-3B-Instruct is the 3B member of Qwen2.5-VL. It accepts text instructions with images or video and generates textual responses, including structured data. Choose it for a clearly defined visual question, a document-extraction task or an explanation that connects visible content with the user’s request. This exact size and release retain their own identity; features added specifically to other sizes or later Qwen generations should not be assumed to apply to this checkpoint.
Visual understanding and agent workflows
Qwen describes recognition of common objects and analysis of writing, charts, icons, graphics and page layouts. For a form, specify the required fields and the meaning of an empty value. For a chart, ask for the axes, units and visible values supporting the conclusion. A focused request makes the output easier to compare with the original image than a broad request to describe everything.
The family also supports visual-agent reasoning and tool direction for computer or phone workflows. The model can interpret an interface and help decide an action; connected application tools perform that action. A screenshot review can ask which control corresponds to a task and what visible information supports the choice. That use of visual reasoning does not itself grant control of a device.
Long video and event localization
The maker describes understanding videos longer than one hour and identifying the segments relevant to a requested event. This is a family capability, not an unconditional media-upload limit or a guarantee that every frame is retained. Prepare the relevant clip, preserve its timing information and ask a precise question about the event, sequence or change. The model’s video-processing approach uses frame-rate information to align visual observations with absolute time.
Localization and structured documents
Qwen2.5-VL can return object locations as bounding boxes or points, with coordinates and attributes in JSON. For a localization task, name the target object and specify whether the result should be a point or a box. Keep the coordinate convention explicit in the consuming application rather than infer a physical measurement from an arbitrary photograph.
The maker also documents structured extraction from scanned invoices, forms and tables. Provide a readable page and an output schema: for example, field name, extracted text and the visible region that supports it. Preserve units, decimal separators and table headings in the source. Treat uncertain or illegible content as a review case rather than asking the model to invent a missing value. These preparation examples apply documented capabilities; they are not guarantees of exact extraction.
Architecture changes
Dynamic image resolution is extended into video through variable frame-rate training. Updated multimodal positional encoding represents temporal sequence and absolute time, helping identify specific moments and changes in speed. The vision encoder uses window attention together with SwiGLU and RMSNorm. These changes describe how the model processes visual material; they do not establish a fixed latency for a particular application.
Preparing images and video
The official examples cover single images, multiple images, text-only conversations, video files and ordered frame lists. Label several visual inputs so a question can refer to each one. For an image comparison, request a table organized by the labels and restrict the result to visible differences. For a video timeline, preserve frame order and timing instead of presenting disconnected pictures as if they carried a known duration.
The image processor’s documented default ranges from four to 16,384 visual tokens per image. Its example can target 256–1,280 tokens by setting pixel bounds. It maintains aspect ratio within those bounds, while explicit resized dimensions are aligned to multiples of 28. More detail consumes more computation and sequence space, so a small text field may need a higher-resolution crop while a broad scene summary may not. Those are maker processing controls; application upload options can differ.
Context and substantive limits
The published checkpoint configuration sets maximum positions to 128,000, while the accompanying long-text section still discusses a 32,768-token starting configuration. The guide describes YaRN extrapolation but warns that it can significantly affect temporal and spatial localization and does not recommend that method for those tasks. It separately notes that video positional encoding can allow a larger position setting, giving 64K as an example. These different configuration descriptions should not be collapsed into a single universally validated input ceiling.
Text, visual material and the answer share a finite sequence budget. Keep relevant content legible, reserve room for the result and check extracted text, coordinates and event times against the original. The card’s 128-token generation example is illustrative and does not establish a separate hard output maximum. This model returns textual analysis or structured information; its vision-language capabilities do not mean it generates images or video.
POST /v1/chat/completionstexttext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen2.5-vl-3b-instruct | Default— |
Fieldmessages | Typemessage[] | RequirementRequired | Values & limits Minimum items: 1 | Default— |
Fieldmessages[].role | Typestring | RequirementRequired | Values & limits Examples: system, user, assistant, tool, function Conditions: These roles are explicitly represented by Source compatibility types; the exact model role set remains model-specific. Model-specific | Default— |
Fieldmessages[].content | Typestring|content_part[]|null | RequirementOptional | Values & limits | Default— |
Fieldmessages[].tool_calls | Typetool_call[] | RequirementOptional | Values & limits | Default— |
Fieldmessages[].tool_call_id | Typestring | RequirementOptional | Values & limits Applies when: role=tool; use actual returned ID | Default— |
Fieldmessages[].name | Typestring | RequirementOptional | Values & limits | Default— |
Fieldmessages[].reasoning_content | Typestring | RequirementOptional | Values & limits Applies when: Round-trip actual returned reasoning where this model uses it | Default— |
Fieldmessages[].reasoning_details | Typeobject|array | RequirementOptional | Values & limits Model-specific | Default— |
Fieldmax_tokens | Typeinteger | RequirementOptional | Values & limits Minimum: 1 Model-specific | Default— |
Fieldmax_completion_tokens | Typeinteger | RequirementOptional | Values & limits Minimum: 1 Model-specific | Default— |
Fieldstream | Typeboolean | RequirementOptional | Values & limits | Defaultfalse |
Fieldstream_options.include_usage | Typeboolean | RequirementOptional | Values & limits Model-specific | Default— |
Fieldtemperature | Typenumber | RequirementOptional | Values & limits Model-specific | Default— |
Fieldtop_p | Typenumber | RequirementOptional | Values & limits Model-specific | Default— |
Fieldstop | Typestring|string[] | RequirementOptional | Values & limits Model-specific | Default— |
Fieldtools | Typefunction_tool[] | RequirementOptional | Values & limits Model-specific | Default— |
Fieldtools[].type | Typestring | RequirementOptional | Values & limits function Model-specific | Default— |
Fieldtools[].function.name | Typestring | RequirementRequired | Values & limits Applies when: function tool supplied | Default— |
Fieldtools[].function.description | Typestring | RequirementOptional | Values & limits | Default— |
Fieldtools[].function.parameters | TypeJSON_schema | RequirementOptional | Values & limits | Default— |
Fieldtool_choice | Typestring|object | RequirementOptional | Values & limits Model-specific | Default— |
Fieldparallel_tool_calls | Typeboolean | RequirementOptional | Values & limits Model-specific | Default— |
Fieldresponse_format | Typeobject | RequirementOptional | Values & limits Model-specific | Default— |
Fieldreasoning_effort | Typestring | RequirementOptional | Values & limits Model-specific | Default— |
Fieldreasoning | Typeobject | RequirementOptional | Values & limits Model-specific | Default— |
Fieldservice_tier | Typestring | RequirementOptional | Values & limits Model-specific | Default— |
Fieldinstructions | Typestring | RequirementOptional | Values & limits Compatibility field | Default— |
Fieldfunctions | Typefunction[] | RequirementOptional | Values & limits Model-specific | Default— |
Fieldfunction_call | Typestring|object | RequirementOptional | Values & limits Model-specific | Default— |
Fieldprompt_cache_options | Typeobject | RequirementOptional | Values & limits Model-specific | Default— |
Fieldmodalities | Typestring[] | RequirementOptional | Values & limits text | Default— |
Fieldprovider | Typeobject | RequirementOptional | Values & limits Conditions: provider.options is rejected on this Ultra text route | Default— |
Fieldcache_control.ttl | Typestring | RequirementOptional | Values & limits Conditions: 1h is rejected in actual prompt-cache locations | Default— |
Text · UTF8 text
messages[].content
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, messages |
Parametermaterial_types | Contracttext |
Parametermessages | Contractmessage array with role/content shape |
Parameterfile_material | Contractactual UTF-8 document inserted as literal text; no file ID inferred |
Parameteroptional_controls | Contractstream, temperature, top_p, stop, max_tokens, max_completion_tokens, reasoning, response_format, tools, tool_choice |
Parametercontrol_scope | Contractmodel-specific; user function tools differ from hosted tools |
Parameteroutput_modalities | Contracttext |
Parameterhosted_operations | Contractfalse |
Parametermultimodal_input_binding | Contractnull |
Lifecycle: synchronous_or_streamed
Send exact model ID with messages.
Context length is not an output cap.
A text file may supply literal text; image/audio/file-object inputs need exact binding.
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 $0.0004 / 1K tokens Output · Output tokens $0.0012 / 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}"\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "qwen2.5-vl-3b-instruct", "messages": [{"role": "user", "content": "Explain the difference between an observation and a hypothesis in three sentences."}]}')\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\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\nr=json.loads(Path("result.json").read_text())\nfor x in r.get("choices",[]):\n if x.get("message",{}).get("content"): print(x["message"]["content"])\nPY_RESULT\nfrom openai import OpenAI\nimport os\n\nif not os.environ.get("AILOOT_API_KEY", "").strip(): raise SystemExit("Set AILOOT_API_KEY before making a request")\nclient=OpenAI(base_url="https://api.ailoot.ai/v1",api_key=os.environ["AILOOT_API_KEY"],max_retries=0)\nresult=client.chat.completions.create(model='qwen2.5-vl-3b-instruct',messages=[{"role":"user","content":"Explain observations and hypotheses in three sentences."}])\nimport base64, json\nfrom pathlib import Path\n\ndef save_result(data, stem="result"):\n # Keep every field, including usage, IDs, reasoning and tool requests.\n Path(stem + ".json").write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")\n print(json.dumps(data, ensure_ascii=False, indent=2))\n def save_blob(encoded, suffix, label):\n raw = base64.b64decode(encoded, altchars=b"-_", validate=True)\n ext = "png" if raw.startswith(b"\\x89PNG\\r\\n\\x1a\\n") else "jpg" if raw.startswith(b"\\xff\\xd8") else "webp" if raw[:4] == b"RIFF" and raw[8:12] == b"WEBP" else suffix\n file = Path(stem + "-" + label + "." + ext)\n file.write_bytes(raw)\n print("Saved", file)\n for index, block in enumerate(data.get("content") or []):\n print("Content", index, json.dumps(block, ensure_ascii=False))\n for choice in data.get("choices") or []:\n print("Choice", choice.get("index"), json.dumps(choice.get("message", choice.get("delta", {})), ensure_ascii=False))\n for index, item in enumerate(data.get("output") or []):\n print("Output", index, json.dumps(item, ensure_ascii=False))\n if item.get("type") == "image_generation_call" and item.get("result"):\n save_blob(item["result"], item.get("output_format") or "bin", "output-" + str(index))\n for ci, candidate in enumerate(data.get("candidates") or []):\n for pi, part in enumerate((candidate.get("content") or {}).get("parts") or []):\n print("Candidate/part", ci, pi, json.dumps(part, ensure_ascii=False))\n blob = part.get("inlineData") or part.get("inline_data")\n if blob and blob.get("data"):\n mime = blob.get("mimeType") or blob.get("mime_type")\n suffix = {"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"}.get(mime, "bin")\n save_blob(blob["data"], suffix, "candidate-" + str(ci) + "-part-" + str(pi))\n\nsave_result(result.to_dict(mode="json"))\nfrom openai import OpenAI\nimport os\nfrom pathlib import Path\n\nif not os.environ.get("AILOOT_API_KEY", "").strip(): raise SystemExit("Set AILOOT_API_KEY before making a request")\nclient=OpenAI(base_url="https://api.ailoot.ai/v1",api_key=os.environ["AILOOT_API_KEY"],max_retries=0)\nimport base64, json\nfrom pathlib import Path\n\ndef save_result(data, stem="result"):\n # Keep every field, including usage, IDs, reasoning and tool requests.\n Path(stem + ".json").write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8")\n print(json.dumps(data, ensure_ascii=False, indent=2))\n def save_blob(encoded, suffix, label):\n raw = base64.b64decode(encoded, altchars=b"-_", validate=True)\n ext = "png" if raw.startswith(b"\\x89PNG\\r\\n\\x1a\\n") else "jpg" if raw.startswith(b"\\xff\\xd8") else "webp" if raw[:4] == b"RIFF" and raw[8:12] == b"WEBP" else suffix\n file = Path(stem + "-" + label + "." + ext)\n file.write_bytes(raw)\n print("Saved", file)\n for index, block in enumerate(data.get("content") or []):\n print("Content", index, json.dumps(block, ensure_ascii=False))\n for choice in data.get("choices") or []:\n print("Choice", choice.get("index"), json.dumps(choice.get("message", choice.get("delta", {})), ensure_ascii=False))\n for index, item in enumerate(data.get("output") or []):\n print("Output", index, json.dumps(item, ensure_ascii=False))\n if item.get("type") == "image_generation_call" and item.get("result"):\n save_blob(item["result"], item.get("output_format") or "bin", "output-" + str(index))\n for ci, candidate in enumerate(data.get("candidates") or []):\n for pi, part in enumerate((candidate.get("content") or {}).get("parts") or []):\n print("Candidate/part", ci, pi, json.dumps(part, ensure_ascii=False))\n blob = part.get("inlineData") or part.get("inline_data")\n if blob and blob.get("data"):\n mime = blob.get("mimeType") or blob.get("mime_type")\n suffix = {"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"}.get(mime, "bin")\n save_blob(blob["data"], suffix, "candidate-" + str(ci) + "-part-" + str(pi))\n\nwith client.chat.completions.stream(model='qwen2.5-vl-3b-instruct',messages=[{"role":"user","content":"Explain observations and hypotheses in three sentences."}]) as stream:\n with Path("stream-events.jsonl").open("w",encoding="utf-8") as log:\n for event in stream:\n payload=event.to_dict(mode="json")\n log.write(json.dumps(payload,ensure_ascii=False)+"\\n"); log.flush()\n print(json.dumps(payload,ensure_ascii=False))\n final=stream.get_final_completion()\n save_result(final.to_dict(mode="json"),"stream-result")\njson or SSE
Keep every complete returned response, file and content part.
Keep every choice and its content/tool calls/reasoning. In SSE assemble deltas by choice/tool index, retain the final finish reason and usage.
Keep request/response IDs and headers. No generic text job retrieval URL is supplied.
| Field | Type |
|---|---|
Fieldchoices[].index | Typeinteger |
Fieldchoices[].message.content | Typestring|null |
Fieldchoices[].message.tool_calls | Typetool_call[] |
Fieldchoices[].message.reasoning_content | Typestring |
Fieldchoices[].message.reasoning_details | Typeobject|array |
Fieldchoices[].finish_reason | Typestring|null |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot Ultra qwen2.5-vl-3b-instruct
Use the method paths shown above.
qwen2.5-vl-3b-instructQwen2.5-VL-3B-Instruct is Qwen’s compact instruction-tuned vision-language model for text, images and video. It returns text and structured visual information, including document fields, object coordinates and event descriptions, with dynamic resolution and time-aware video processing.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Introduction: the instruction-tuned 3B edition
Qwen2.5-VL-3B-Instruct is the 3B member of Qwen2.5-VL. It accepts text instructions with images or video and generates textual responses, including structured data. Choose it for a clearly defined visual question, a document-extraction task or an explanation that connects visible content with the user’s request. This exact size and release retain their own identity; features added specifically to other sizes or later Qwen generations should not be assumed to apply to this checkpoint.
Visual understanding and agent workflows
Qwen describes recognition of common objects and analysis of writing, charts, icons, graphics and page layouts. For a form, specify the required fields and the meaning of an empty value. For a chart, ask for the axes, units and visible values supporting the conclusion. A focused request makes the output easier to compare with the original image than a broad request to describe everything.
The family also supports visual-agent reasoning and tool direction for computer or phone workflows. The model can interpret an interface and help decide an action; connected application tools perform that action. A screenshot review can ask which control corresponds to a task and what visible information supports the choice. That use of visual reasoning does not itself grant control of a device.
Long video and event localization
The maker describes understanding videos longer than one hour and identifying the segments relevant to a requested event. This is a family capability, not an unconditional media-upload limit or a guarantee that every frame is retained. Prepare the relevant clip, preserve its timing information and ask a precise question about the event, sequence or change. The model’s video-processing approach uses frame-rate information to align visual observations with absolute time.
Localization and structured documents
Qwen2.5-VL can return object locations as bounding boxes or points, with coordinates and attributes in JSON. For a localization task, name the target object and specify whether the result should be a point or a box. Keep the coordinate convention explicit in the consuming application rather than infer a physical measurement from an arbitrary photograph.
The maker also documents structured extraction from scanned invoices, forms and tables. Provide a readable page and an output schema: for example, field name, extracted text and the visible region that supports it. Preserve units, decimal separators and table headings in the source. Treat uncertain or illegible content as a review case rather than asking the model to invent a missing value. These preparation examples apply documented capabilities; they are not guarantees of exact extraction.
Architecture changes
Dynamic image resolution is extended into video through variable frame-rate training. Updated multimodal positional encoding represents temporal sequence and absolute time, helping identify specific moments and changes in speed. The vision encoder uses window attention together with SwiGLU and RMSNorm. These changes describe how the model processes visual material; they do not establish a fixed latency for a particular application.
Preparing images and video
The official examples cover single images, multiple images, text-only conversations, video files and ordered frame lists. Label several visual inputs so a question can refer to each one. For an image comparison, request a table organized by the labels and restrict the result to visible differences. For a video timeline, preserve frame order and timing instead of presenting disconnected pictures as if they carried a known duration.
The image processor’s documented default ranges from four to 16,384 visual tokens per image. Its example can target 256–1,280 tokens by setting pixel bounds. It maintains aspect ratio within those bounds, while explicit resized dimensions are aligned to multiples of 28. More detail consumes more computation and sequence space, so a small text field may need a higher-resolution crop while a broad scene summary may not. Those are maker processing controls; application upload options can differ.
Context and substantive limits
The published checkpoint configuration sets maximum positions to 128,000, while the accompanying long-text section still discusses a 32,768-token starting configuration. The guide describes YaRN extrapolation but warns that it can significantly affect temporal and spatial localization and does not recommend that method for those tasks. It separately notes that video positional encoding can allow a larger position setting, giving 64K as an example. These different configuration descriptions should not be collapsed into a single universally validated input ceiling.
Text, visual material and the answer share a finite sequence budget. Keep relevant content legible, reserve room for the result and check extracted text, coordinates and event times against the original. The card’s 128-token generation example is illustrative and does not establish a separate hard output maximum. This model returns textual analysis or structured information; its vision-language capabilities do not mean it generates images or video.
POST /v1/messagestextjson, sse| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen2.5-vl-3b-instruct | Default— |
Fieldmessages | Typemessage[] | RequirementRequired | Values & limits | Default— |
Fieldmessages[].role | Typestring | RequirementRequired | Values & limits user, assistant | Default— |
Fieldmessages[].content | Typestring | RequirementRequired | Values & limits | Default— |
Fieldmax_tokens | Typeinteger | RequirementRequired | Values & limits | Default— |
Fieldstream | Typeboolean | RequirementOptional | Values & limits | Defaultfalse |
Text · UTF-8 text
messages[].content string
| Parameter | Contract |
|---|---|
Parametercontent_type | Contractapplication/json |
Parameterrequired | Contractmodel, messages, max_tokens |
Parameteroptional_controls | Contractstream |
Parametermaterial_types | Contracttext |
Parametertext_file_material | ContractRead the actual UTF-8 document and put its text in the indicated input or message content string. |
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 $0.0004 / 1K tokens Output · Output tokens $0.0012 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
{\n "model": "qwen2.5-vl-3b-instruct",\n "max_tokens": 128,\n "messages": [\n {\n "role": "user",\n "content": "Summarize the incident report."\n }\n ],\n "stream": false\n}set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 document}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json, os\nfrom pathlib import Path\n\nsource_file = Path(os.environ['AILOOT_TEXT_FILE'])\nif not source_file.is_file():\n raise SystemExit('Set AILOOT_TEXT_FILE to your actual UTF-8 document')\ntext = source_file.read_text(encoding='utf-8')\nif not text.strip():\n raise SystemExit('Choose a document containing text')\npayload = {'model': 'qwen2.5-vl-3b-instruct', 'max_tokens': 128, 'messages': [{'role': 'user', 'content': 'Summarize the incident report.'}]}\npayload['messages'][0]['content'] = 'Summarize this material and quote supporting source passages.\\n\\n' + text\nPath('request.json').write_text(json.dumps(payload, ensure_ascii=False), encoding='utf-8')\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/messages \\\n -H "x-api-key: $AILOOT_API_KEY" -H 'anthropic-version: 2023-06-01' -H 'Content-Type: application/json' \\\n --data-binary @request.json -D response-headers.txt -o result.json\npython3 - <<'PY_RESULT'\nimport json\nfrom pathlib import Path\n\nresult = json.loads(Path('result.json').read_text(encoding='utf-8'))\nprint(json.dumps(result, ensure_ascii=False, indent=2))\nPY_RESULT\nimport json, os\nfrom pathlib import Path\n\nsource_file = Path(os.environ['AILOOT_TEXT_FILE'])\nif not source_file.is_file():\n raise SystemExit('Set AILOOT_TEXT_FILE to your actual UTF-8 document')\ntext = source_file.read_text(encoding='utf-8')\nif not text.strip():\n raise SystemExit('Choose a document containing text')\npayload = {'model': 'qwen2.5-vl-3b-instruct', 'max_tokens': 128, 'messages': [{'role': 'user', 'content': 'Summarize the incident report.'}]}\npayload['messages'][0]['content'] = 'Summarize this material and quote supporting source passages.\\n\\n' + text\nfrom urllib.request import Request, urlopen\nfrom urllib.error import HTTPError\n\napi_key = os.environ['AILOOT_API_KEY']\nif not api_key.strip():\n raise SystemExit('Set AILOOT_API_KEY before making a request')\nheaders = {'x-api-key': 'PLACEHOLDER_KEY', 'anthropic-version': '2023-06-01', 'Content-Type': 'application/json'}\nheaders['x-api-key'] = api_key\nrequest = Request('https://api.ailoot.ai/v1/messages', data=json.dumps(payload, ensure_ascii=False).encode('utf-8'), headers=headers, method='POST')\ntry:\n with urlopen(request, timeout=120) as response:\n raw = response.read()\n Path('response-headers.txt').write_text(str(response.headers), encoding='utf-8')\nexcept HTTPError as error:\n Path('result.json').write_bytes(error.read())\n Path('response-headers.txt').write_text(str(error.headers), encoding='utf-8')\n raise\nPath('result.json').write_bytes(raw)\nimport json\nfrom pathlib import Path\n\nresult = json.loads(Path('result.json').read_text(encoding='utf-8'))\nprint(json.dumps(result, ensure_ascii=False, indent=2))\nJSON or SSE
Keep every complete returned response, file and content part.
Keep all content blocks, thinking signatures and actual tool-use IDs. Assemble all SSE block deltas; a first text block is not the complete output.
Keep request/response IDs and headers. No generic text job retrieval URL is supplied.
| Field | Type |
|---|---|
Fieldid | Typestring |
Fieldmodel | Typestring |
Fieldcontent[] | Typetext|thinking|tool_use|other content block |
Fieldstop_reason | Typestring|null |
Fieldusage | Typeobject |
| Response | Condition | Next step |
|---|---|---|
Response400 · invalid_request_error | ConditionInvalid required fields, material representation or parameter value | Next stepCorrect the indicated fields before a new creation request. |
Ailoot Ultra qwen2.5-vl-3b-instruct
Use the method paths shown above.
Technical example in English
Organize a recorded event timeline\n\nUse only these text annotations from a clip: 00:02 worker picks up a box; 00:05 worker places the box on the scale; 00:08 scale display is obscured; 00:11 worker removes the box. Return the event sequence, the first annotated placement time and whether the weight is established. Preserve timestamps without claiming frame-level inspection.An ordered event table, a placement-time field and an unresolved weight field.Check that obscured weight remains unknown and the placement timestamp is attributed to the supplied annotation.