Qwen3-4B applies Qwen3’s hybrid response design to a compact dense language model. The manufacturer describes mathematical and logical reasoning, code generation, creative writing, role-based interaction and multi-turn dialogue, alongside instruction following in more than 100 languages and dialects. A focused editorial or technical task is a useful way to work with it: provide a specific source fragment and ask for one well-defined transformation instead of an open-ended overview.
The release contains four billion parameters, with 3.6 billion outside embeddings, and has 36 layers. Grouped-query attention uses 32 query heads and eight key/value heads. Text is both the input and output. This card describes the original Qwen3-4B hybrid release, not a later dated Thinking or Instruct model. A practical example is to turn a short incident note into an operational handoff containing the observed failure, known impact and next investigation step. Supply those facts as separate fields and require the model to mark missing information rather than fill it in.
Prepare the instruction and source as distinct parts of a chat message. For programming, attach a small function, representative input and the desired return value, then request an explanation or a narrowly scoped change. Ask for an ordinary case and an edge case as part of the output. For a writing task, state the audience, length and vocabulary that must remain unchanged. For a classification task, provide the allowed labels, their definitions and a sample output; the returned choice should then be checked against that list. The published integration expects a Qwen3-aware runtime; the card warns that Transformers versions below 4.51.0 do not recognize the required architecture.
Thinking and non-thinking are built into the same release. In the manufacturer’s template, thinking is on by default, and a hard switch turns it off. With thinking enabled, /think and /no_think are soft instructions that can change the behavior between turns; the latest instruction applies. They do not reactivate a hard-disabled mode. A short text transformation can request a direct answer, while an explanation of conflicting conditions can use reasoning. Store the final response, rather than prior reasoning text, as the useful result of a conversation turn. With thinking enabled, even a soft non-thinking request retains a reasoning block; its content may be empty. The hard disable removes that block.
The model card includes agentic use and recommends Qwen-Agent for tool definitions and parsing. For an assistant that classifies an incoming ticket and then looks up a category description, the application would expose the lookup tool, execute any accepted call and supply the result. Define the arguments and expected fields beforehand. Request a final ticket record with the selected category, a short justification and any unknown fields. Validate the record and retain the tool result used to create it. The Qwen3 guide also describes parallel and multi-step tool calls, with arguments represented as objects and results supplied back into the conversation.
The exact card gives 32,768 native context tokens and documents extension to 131,072 using YaRN. The expanded figure is tied to configuration; static scaling may affect shorter inputs. Plan the full request and response together. For a collection of short notes, label each item, ask for a matching output identifier and use a concise schema so the generated result still fits. The card’s output-length examples and recommendations are not a universal hard generation limit. The Qwen3 documentation discusses assistant-message lengths of 38,912 tokens in thinking mode and 16,384 in non-thinking mode in its training-length explanation, and recommends evaluating the quality of longer generation. These figures are distinct from the overall context window and are not an endpoint-specific hard cap. The card also explains its default positional allocation of 40,960 tokens as 8,192 for typical prompts plus 32,768 for output; this is manufacturer configuration guidance, separate from the native training context and any endpoint limit.
This edition has an additional warning about persistent repetition. The manufacturer points to the recommended sampling setup and suggests presence_penalty 1.5 if significant repetition occurs. Its general advice notes that stronger penalties can introduce mixed-language output or reduce quality, so this is a targeted setting rather than a guarantee. Thinking mode uses temperature 0.6, top-p 0.95, top-k 20 and min-p 0; non-thinking uses 0.7, 0.8, 20 and 0. Avoid greedy decoding for reasoning, require a clearly delimited final answer and inspect the result for repeated clauses, missing fields and changed terminology.
", "aboutText": "Qwen3-4B applies Qwen3’s hybrid response design to a compact dense language model. The manufacturer describes mathematical and logical reasoning, code generation, creative writing, role-based interaction and multi-turn dialogue, alongside instruction following in more than 100 languages and dialects. A focused editorial or technical task is a useful way to work with it: provide a specific source fragment and ask for one well-defined transformation instead of an open-ended overview.\n\nThe release contains four billion parameters, with 3.6 billion outside embeddings, and has 36 layers. Grouped-query attention uses 32 query heads and eight key/value heads. Text is both the input and output. This card describes the original Qwen3-4B hybrid release, not a later dated Thinking or Instruct model. A practical example is to turn a short incident note into an operational handoff containing the observed failure, known impact and next investigation step. Supply those facts as separate fields and require the model to mark missing information rather than fill it in.\n\nPrepare the instruction and source as distinct parts of a chat message. For programming, attach a small function, representative input and the desired return value, then request an explanation or a narrowly scoped change. Ask for an ordinary case and an edge case as part of the output. For a writing task, state the audience, length and vocabulary that must remain unchanged. For a classification task, provide the allowed labels, their definitions and a sample output; the returned choice should then be checked against that list. The published integration expects a Qwen3-aware runtime; the card warns that Transformers versions below 4.51.0 do not recognize the required architecture.\n\nThinking and non-thinking are built into the same release. In the manufacturer’s template, thinking is on by default, and a hard switch turns it off. With thinking enabled, /think and /no_think are soft instructions that can change the behavior between turns; the latest instruction applies. They do not reactivate a hard-disabled mode. A short text transformation can request a direct answer, while an explanation of conflicting conditions can use reasoning. Store the final response, rather than prior reasoning text, as the useful result of a conversation turn. With thinking enabled, even a soft non-thinking request retains a reasoning block; its content may be empty. The hard disable removes that block.\n\nThe model card includes agentic use and recommends Qwen-Agent for tool definitions and parsing. For an assistant that classifies an incoming ticket and then looks up a category description, the application would expose the lookup tool, execute any accepted call and supply the result. Define the arguments and expected fields beforehand. Request a final ticket record with the selected category, a short justification and any unknown fields. Validate the record and retain the tool result used to create it. The Qwen3 guide also describes parallel and multi-step tool calls, with arguments represented as objects and results supplied back into the conversation.\n\nThe exact card gives 32,768 native context tokens and documents extension to 131,072 using YaRN. The expanded figure is tied to configuration; static scaling may affect shorter inputs. Plan the full request and response together. For a collection of short notes, label each item, ask for a matching output identifier and use a concise schema so the generated result still fits. The card’s output-length examples and recommendations are not a universal hard generation limit. The Qwen3 documentation discusses assistant-message lengths of 38,912 tokens in thinking mode and 16,384 in non-thinking mode in its training-length explanation, and recommends evaluating the quality of longer generation. These figures are distinct from the overall context window and are not an endpoint-specific hard cap. The card also explains its default positional allocation of 40,960 tokens as 8,192 for typical prompts plus 32,768 for output; this is manufacturer configuration guidance, separate from the native training context and any endpoint limit.\n\nThis edition has an additional warning about persistent repetition. The manufacturer points to the recommended sampling setup and suggests presence_penalty 1.5 if significant repetition occurs. Its general advice notes that stronger penalties can introduce mixed-language output or reduce quality, so this is a targeted setting rather than a guarantee. Thinking mode uses temperature 0.6, top-p 0.95, top-k 20 and min-p 0; non-thinking uses 0.7, 0.8, 20 and 0. Avoid greedy decoding for reasoning, require a clearly delimited final answer and inspect the result for repeated clauses, missing fields and changed terminology.", "editorialOverlay": { "copy_fields": { "practical_html": "Technical example in English
Turn an incident note into a handoff\n\nTurn these facts into an operational handoff. Observed failure: report export stops with Disk full. Known impact: exports for Team C failed; other teams were not checked. Last completed step: temporary files listed, none deleted. Next step: identify approved cleanup candidates. Use four labelled lines and do not invent a restoration time or wider impact.Four labelled lines: Failure, Confirmed impact, Completed work and Next step.Check that listing files is not reported as deleting them and untested teams remain outside confirmed impact.qwen3-4bQwen3-4B is a four-billion-parameter text model for instruction-following, multilingual dialogue and code. Its hybrid design supports both direct answers and a thinking mode, with edition-specific guidance for controlling repetition.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen3-4B applies Qwen3’s hybrid response design to a compact dense language model. The manufacturer describes mathematical and logical reasoning, code generation, creative writing, role-based interaction and multi-turn dialogue, alongside instruction following in more than 100 languages and dialects. A focused editorial or technical task is a useful way to work with it: provide a specific source fragment and ask for one well-defined transformation instead of an open-ended overview.
The release contains four billion parameters, with 3.6 billion outside embeddings, and has 36 layers. Grouped-query attention uses 32 query heads and eight key/value heads. Text is both the input and output. This card describes the original Qwen3-4B hybrid release, not a later dated Thinking or Instruct model. A practical example is to turn a short incident note into an operational handoff containing the observed failure, known impact and next investigation step. Supply those facts as separate fields and require the model to mark missing information rather than fill it in.
Prepare the instruction and source as distinct parts of a chat message. For programming, attach a small function, representative input and the desired return value, then request an explanation or a narrowly scoped change. Ask for an ordinary case and an edge case as part of the output. For a writing task, state the audience, length and vocabulary that must remain unchanged. For a classification task, provide the allowed labels, their definitions and a sample output; the returned choice should then be checked against that list. The published integration expects a Qwen3-aware runtime; the card warns that Transformers versions below 4.51.0 do not recognize the required architecture.
Thinking and non-thinking are built into the same release. In the manufacturer’s template, thinking is on by default, and a hard switch turns it off. With thinking enabled, /think and /no_think are soft instructions that can change the behavior between turns; the latest instruction applies. They do not reactivate a hard-disabled mode. A short text transformation can request a direct answer, while an explanation of conflicting conditions can use reasoning. Store the final response, rather than prior reasoning text, as the useful result of a conversation turn. With thinking enabled, even a soft non-thinking request retains a reasoning block; its content may be empty. The hard disable removes that block.
The model card includes agentic use and recommends Qwen-Agent for tool definitions and parsing. For an assistant that classifies an incoming ticket and then looks up a category description, the application would expose the lookup tool, execute any accepted call and supply the result. Define the arguments and expected fields beforehand. Request a final ticket record with the selected category, a short justification and any unknown fields. Validate the record and retain the tool result used to create it. The Qwen3 guide also describes parallel and multi-step tool calls, with arguments represented as objects and results supplied back into the conversation.
The exact card gives 32,768 native context tokens and documents extension to 131,072 using YaRN. The expanded figure is tied to configuration; static scaling may affect shorter inputs. Plan the full request and response together. For a collection of short notes, label each item, ask for a matching output identifier and use a concise schema so the generated result still fits. The card’s output-length examples and recommendations are not a universal hard generation limit. The Qwen3 documentation discusses assistant-message lengths of 38,912 tokens in thinking mode and 16,384 in non-thinking mode in its training-length explanation, and recommends evaluating the quality of longer generation. These figures are distinct from the overall context window and are not an endpoint-specific hard cap. The card also explains its default positional allocation of 40,960 tokens as 8,192 for typical prompts plus 32,768 for output; this is manufacturer configuration guidance, separate from the native training context and any endpoint limit.
This edition has an additional warning about persistent repetition. The manufacturer points to the recommended sampling setup and suggests presence_penalty 1.5 if significant repetition occurs. Its general advice notes that stronger penalties can introduce mixed-language output or reduce quality, so this is a targeted setting rather than a guarantee. Thinking mode uses temperature 0.6, top-p 0.95, top-k 20 and min-p 0; non-thinking uses 0.7, 0.8, 20 and 0. Avoid greedy decoding for reasoning, require a clearly delimited final answer and inspect the result for repeated clauses, missing fields and changed terminology.
POST /v1/chat/completionstexttext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen3-4b | 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 $0.0001 / 1K tokens Output · Output tokens $0.001 / 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": "qwen3-4b", "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='qwen3-4b',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='qwen3-4b',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 qwen3-4b
Use the method paths shown above.
qwen3-4bQwen3-4B is a four-billion-parameter text model for instruction-following, multilingual dialogue and code. Its hybrid design supports both direct answers and a thinking mode, with edition-specific guidance for controlling repetition.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen3-4B applies Qwen3’s hybrid response design to a compact dense language model. The manufacturer describes mathematical and logical reasoning, code generation, creative writing, role-based interaction and multi-turn dialogue, alongside instruction following in more than 100 languages and dialects. A focused editorial or technical task is a useful way to work with it: provide a specific source fragment and ask for one well-defined transformation instead of an open-ended overview.
The release contains four billion parameters, with 3.6 billion outside embeddings, and has 36 layers. Grouped-query attention uses 32 query heads and eight key/value heads. Text is both the input and output. This card describes the original Qwen3-4B hybrid release, not a later dated Thinking or Instruct model. A practical example is to turn a short incident note into an operational handoff containing the observed failure, known impact and next investigation step. Supply those facts as separate fields and require the model to mark missing information rather than fill it in.
Prepare the instruction and source as distinct parts of a chat message. For programming, attach a small function, representative input and the desired return value, then request an explanation or a narrowly scoped change. Ask for an ordinary case and an edge case as part of the output. For a writing task, state the audience, length and vocabulary that must remain unchanged. For a classification task, provide the allowed labels, their definitions and a sample output; the returned choice should then be checked against that list. The published integration expects a Qwen3-aware runtime; the card warns that Transformers versions below 4.51.0 do not recognize the required architecture.
Thinking and non-thinking are built into the same release. In the manufacturer’s template, thinking is on by default, and a hard switch turns it off. With thinking enabled, /think and /no_think are soft instructions that can change the behavior between turns; the latest instruction applies. They do not reactivate a hard-disabled mode. A short text transformation can request a direct answer, while an explanation of conflicting conditions can use reasoning. Store the final response, rather than prior reasoning text, as the useful result of a conversation turn. With thinking enabled, even a soft non-thinking request retains a reasoning block; its content may be empty. The hard disable removes that block.
The model card includes agentic use and recommends Qwen-Agent for tool definitions and parsing. For an assistant that classifies an incoming ticket and then looks up a category description, the application would expose the lookup tool, execute any accepted call and supply the result. Define the arguments and expected fields beforehand. Request a final ticket record with the selected category, a short justification and any unknown fields. Validate the record and retain the tool result used to create it. The Qwen3 guide also describes parallel and multi-step tool calls, with arguments represented as objects and results supplied back into the conversation.
The exact card gives 32,768 native context tokens and documents extension to 131,072 using YaRN. The expanded figure is tied to configuration; static scaling may affect shorter inputs. Plan the full request and response together. For a collection of short notes, label each item, ask for a matching output identifier and use a concise schema so the generated result still fits. The card’s output-length examples and recommendations are not a universal hard generation limit. The Qwen3 documentation discusses assistant-message lengths of 38,912 tokens in thinking mode and 16,384 in non-thinking mode in its training-length explanation, and recommends evaluating the quality of longer generation. These figures are distinct from the overall context window and are not an endpoint-specific hard cap. The card also explains its default positional allocation of 40,960 tokens as 8,192 for typical prompts plus 32,768 for output; this is manufacturer configuration guidance, separate from the native training context and any endpoint limit.
This edition has an additional warning about persistent repetition. The manufacturer points to the recommended sampling setup and suggests presence_penalty 1.5 if significant repetition occurs. Its general advice notes that stronger penalties can introduce mixed-language output or reduce quality, so this is a targeted setting rather than a guarantee. Thinking mode uses temperature 0.6, top-p 0.95, top-k 20 and min-p 0; non-thinking uses 0.7, 0.8, 20 and 0. Avoid greedy decoding for reasoning, require a clearly delimited final answer and inspect the result for repeated clauses, missing fields and changed terminology.
POST /v1/messagestextjson, sse| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen3-4b | 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 $0.0001 / 1K tokens Output · Output tokens $0.001 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
{\n "model": "qwen3-4b",\n "max_tokens": 128,\n "messages": [\n {\n "role": "user",\n "content": "Summarize the incident report."\n }\n ],\n "stream": false\n}set -eu\n: "${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': 'qwen3-4b', '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': 'qwen3-4b', '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 qwen3-4b
Use the method paths shown above.
Technical example in English
Turn an incident note into a handoff\n\nTurn these facts into an operational handoff. Observed failure: report export stops with Disk full. Known impact: exports for Team C failed; other teams were not checked. Last completed step: temporary files listed, none deleted. Next step: identify approved cleanup candidates. Use four labelled lines and do not invent a restoration time or wider impact.Four labelled lines: Failure, Confirmed impact, Completed work and Next step.Check that listing files is not reported as deleting them and untested teams remain outside confirmed impact.