Qwen3.5 highlights: a sparse vision-language checkpoint
Qwen3.5-35B-A3B is a post-trained vision-language model combining text, image and video understanding with textual responses. The family is trained with early text–vision fusion and reinforcement learning across progressively complex tasks, supporting reasoning, coding, agents and visual analysis. It also describes coverage of 201 languages and dialects. That coverage is a family capability rather than a promise of equal quality for every language or task.
The exact open checkpoint is separate from Qwen3.5-35B-A3B-Base and the later Qwen3.6 release. The maker also describes a managed Flash offering with additional features; those interface features should not be assigned automatically to this open checkpoint. Preserve the requested 35B-A3B identity when comparing results or retaining a reproducible model selection.
Model overview
The mixture-of-experts model has 35 billion parameters in total and activates three billion. Its 40-layer hybrid layout repeats three Gated DeltaNet expert layers followed by one gated-attention expert layer. It has 256 routed experts, with eight routed experts and one shared expert active in the documented layout. A vision encoder and multi-token prediction are part of the model. The activated count does not describe the complete checkpoint size, and architecture alone does not guarantee latency in a particular application.
The native context is 262,144 tokens, with conditional extension to 1,010,000. Qwen advises retaining at least a 128K context where possible for complex thinking tasks, even if a smaller window would reduce memory pressure. Text, represented visual material, history, reasoning and the final answer all need space in the configured sequence.
Using inputs and response modes
For a screenshot-informed coding request, provide the visual reference, existing source context, interaction requirements and expected editable artifact. For a chart or document, retain labels and units and request a table of values with supporting evidence. For a multilingual explanation, state the target language and terms to preserve.
Thinking is enabled by default, producing a reasoning block before the final response. The maker’s enable_thinking control can disable it for a direct answer. Qwen3.5 does not officially support the old /think and /nothink soft prompt switches. Its history guidance keeps final answers without thinking content. This should remain distinct from later Qwen3.6 preservation options or Qwen3.8’s default historical reasoning retention.
Agentic usage
The maker documents tool calling and recommends Qwen-Agent with configured or MCP-connected tools, alongside Qwen Code for repository-oriented work. Define the permitted operations, relevant evidence and final deliverable. The application executes tools and supplies their observations; the model can use that feedback to revise its analysis. A planned command or generated patch is not an executed action, and the model name alone does not establish access to current information or files.
Ultra-long context
The maker’s extension recipe uses YaRN with the checkpoint’s interleaved positional settings. A factor of four targets the roughly one-million-token workflow; a factor of two is suggested for 524,288-token tasks. Static scaling can reduce shorter-text performance, so it should be configured when the larger context is needed. Organize long material with document boundaries, code identifiers and a clear question instead of filling the window with unrelated evidence.
Generation recommendations
In Best Practices, general thinking uses temperature 1.0, top_p 0.95, top_k 20, min_p 0, presence penalty 1.5 and repetition penalty 1.0. Precise coding uses temperature 0.6 and presence penalty zero. General direct responses use temperature 0.7, top_p 0.8 and presence penalty 1.5; top_k is 20, min_p is 0 and repetition penalty is 1.0. For reasoning in direct mode, Best Practices give temperature 1.0, top_p 1.0, top_k 40 and presence penalty 2.0. Keep these task-specific presets distinct; larger presence penalties can cause language mixing or a slight quality decrease.
Recommended generation allowances are 32,768 tokens for most questions and 81,920 for unusually complex mathematics or programming. These are not separately established hard output ceilings. Define the final-answer form explicitly and check calculations or generated code against the supplied conditions.
Video and substantive limits
The supported video example defaults to two frames per second. The maker’s hour-scale video guidance expands the processor budget to roughly 224K video tokens for higher-rate sampling. That requires configuration and consumes context alongside the task and answer; it is not an unconditional duration or perfect-event-recall guarantee. Preserve relevant scene transitions and small writing, and request a result tied to the visible evidence. The practical value is the checkpoint’s joint text/visual scope and selectable response mode, with a concrete artifact that can be reviewed.
", "aboutText": "Qwen3.5 highlights: a sparse vision-language checkpoint\n\nQwen3.5-35B-A3B is a post-trained vision-language model combining text, image and video understanding with textual responses. The family is trained with early text–vision fusion and reinforcement learning across progressively complex tasks, supporting reasoning, coding, agents and visual analysis. It also describes coverage of 201 languages and dialects. That coverage is a family capability rather than a promise of equal quality for every language or task.\n\nThe exact open checkpoint is separate from Qwen3.5-35B-A3B-Base and the later Qwen3.6 release. The maker also describes a managed Flash offering with additional features; those interface features should not be assigned automatically to this open checkpoint. Preserve the requested 35B-A3B identity when comparing results or retaining a reproducible model selection.\n\nModel overview\n\nThe mixture-of-experts model has 35 billion parameters in total and activates three billion. Its 40-layer hybrid layout repeats three Gated DeltaNet expert layers followed by one gated-attention expert layer. It has 256 routed experts, with eight routed experts and one shared expert active in the documented layout. A vision encoder and multi-token prediction are part of the model. The activated count does not describe the complete checkpoint size, and architecture alone does not guarantee latency in a particular application.\n\nThe native context is 262,144 tokens, with conditional extension to 1,010,000. Qwen advises retaining at least a 128K context where possible for complex thinking tasks, even if a smaller window would reduce memory pressure. Text, represented visual material, history, reasoning and the final answer all need space in the configured sequence.\n\nUsing inputs and response modes\n\nFor a screenshot-informed coding request, provide the visual reference, existing source context, interaction requirements and expected editable artifact. For a chart or document, retain labels and units and request a table of values with supporting evidence. For a multilingual explanation, state the target language and terms to preserve.\n\nThinking is enabled by default, producing a reasoning block before the final response. The maker’s enable_thinking control can disable it for a direct answer. Qwen3.5 does not officially support the old /think and /nothink soft prompt switches. Its history guidance keeps final answers without thinking content. This should remain distinct from later Qwen3.6 preservation options or Qwen3.8’s default historical reasoning retention.\n\nAgentic usage\n\nThe maker documents tool calling and recommends Qwen-Agent with configured or MCP-connected tools, alongside Qwen Code for repository-oriented work. Define the permitted operations, relevant evidence and final deliverable. The application executes tools and supplies their observations; the model can use that feedback to revise its analysis. A planned command or generated patch is not an executed action, and the model name alone does not establish access to current information or files.\n\nUltra-long context\n\nThe maker’s extension recipe uses YaRN with the checkpoint’s interleaved positional settings. A factor of four targets the roughly one-million-token workflow; a factor of two is suggested for 524,288-token tasks. Static scaling can reduce shorter-text performance, so it should be configured when the larger context is needed. Organize long material with document boundaries, code identifiers and a clear question instead of filling the window with unrelated evidence.\n\nGeneration recommendations\n\nIn Best Practices, general thinking uses temperature 1.0, top_p 0.95, top_k 20, min_p 0, presence penalty 1.5 and repetition penalty 1.0. Precise coding uses temperature 0.6 and presence penalty zero. General direct responses use temperature 0.7, top_p 0.8 and presence penalty 1.5; top_k is 20, min_p is 0 and repetition penalty is 1.0. For reasoning in direct mode, Best Practices give temperature 1.0, top_p 1.0, top_k 40 and presence penalty 2.0. Keep these task-specific presets distinct; larger presence penalties can cause language mixing or a slight quality decrease.\n\nRecommended generation allowances are 32,768 tokens for most questions and 81,920 for unusually complex mathematics or programming. These are not separately established hard output ceilings. Define the final-answer form explicitly and check calculations or generated code against the supplied conditions.\n\nVideo and substantive limits\n\nThe supported video example defaults to two frames per second. The maker’s hour-scale video guidance expands the processor budget to roughly 224K video tokens for higher-rate sampling. That requires configuration and consumes context alongside the task and answer; it is not an unconditional duration or perfect-event-recall guarantee. Preserve relevant scene transitions and small writing, and request a result tied to the visible evidence. The practical value is the checkpoint’s joint text/visual scope and selectable response mode, with a concrete artifact that can be reviewed.", "editorialOverlay": { "copy_fields": { "practical_html": "Technical example in English
Audit a transcribed experiment table\n\nUse this transcribed experiment table: baseline, 100 trials, 12 failures; revision A, 80 trials, 8 failures; revision B, 40 trials, 2 failures. Calculate the observed failure rates and compare them. Explain what these counts establish and what would require additional statistical analysis; do not claim that the smallest observed rate proves the best underlying system.A rate table and a qualified comparison separating observed proportions from unperformed statistical inference.Each rate uses its own trial count, and the conclusion does not treat unequal sample sizes as equivalent proof.qwen/qwen3.5-35b-a3bQwen3.5-35B-A3B is Qwen’s vision-language mixture-of-experts model with 35 billion total and 3 billion activated parameters. It accepts text, images and video for reasoning, coding and tool-assisted work, with thinking and direct-answer modes and a native 262,144-token context.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen3.5 highlights: a sparse vision-language checkpoint
Qwen3.5-35B-A3B is a post-trained vision-language model combining text, image and video understanding with textual responses. The family is trained with early text–vision fusion and reinforcement learning across progressively complex tasks, supporting reasoning, coding, agents and visual analysis. It also describes coverage of 201 languages and dialects. That coverage is a family capability rather than a promise of equal quality for every language or task.
The exact open checkpoint is separate from Qwen3.5-35B-A3B-Base and the later Qwen3.6 release. The maker also describes a managed Flash offering with additional features; those interface features should not be assigned automatically to this open checkpoint. Preserve the requested 35B-A3B identity when comparing results or retaining a reproducible model selection.
Model overview
The mixture-of-experts model has 35 billion parameters in total and activates three billion. Its 40-layer hybrid layout repeats three Gated DeltaNet expert layers followed by one gated-attention expert layer. It has 256 routed experts, with eight routed experts and one shared expert active in the documented layout. A vision encoder and multi-token prediction are part of the model. The activated count does not describe the complete checkpoint size, and architecture alone does not guarantee latency in a particular application.
The native context is 262,144 tokens, with conditional extension to 1,010,000. Qwen advises retaining at least a 128K context where possible for complex thinking tasks, even if a smaller window would reduce memory pressure. Text, represented visual material, history, reasoning and the final answer all need space in the configured sequence.
Using inputs and response modes
For a screenshot-informed coding request, provide the visual reference, existing source context, interaction requirements and expected editable artifact. For a chart or document, retain labels and units and request a table of values with supporting evidence. For a multilingual explanation, state the target language and terms to preserve.
Thinking is enabled by default, producing a reasoning block before the final response. The maker’s enable_thinking control can disable it for a direct answer. Qwen3.5 does not officially support the old /think and /nothink soft prompt switches. Its history guidance keeps final answers without thinking content. This should remain distinct from later Qwen3.6 preservation options or Qwen3.8’s default historical reasoning retention.
Agentic usage
The maker documents tool calling and recommends Qwen-Agent with configured or MCP-connected tools, alongside Qwen Code for repository-oriented work. Define the permitted operations, relevant evidence and final deliverable. The application executes tools and supplies their observations; the model can use that feedback to revise its analysis. A planned command or generated patch is not an executed action, and the model name alone does not establish access to current information or files.
Ultra-long context
The maker’s extension recipe uses YaRN with the checkpoint’s interleaved positional settings. A factor of four targets the roughly one-million-token workflow; a factor of two is suggested for 524,288-token tasks. Static scaling can reduce shorter-text performance, so it should be configured when the larger context is needed. Organize long material with document boundaries, code identifiers and a clear question instead of filling the window with unrelated evidence.
Generation recommendations
In Best Practices, general thinking uses temperature 1.0, top_p 0.95, top_k 20, min_p 0, presence penalty 1.5 and repetition penalty 1.0. Precise coding uses temperature 0.6 and presence penalty zero. General direct responses use temperature 0.7, top_p 0.8 and presence penalty 1.5; top_k is 20, min_p is 0 and repetition penalty is 1.0. For reasoning in direct mode, Best Practices give temperature 1.0, top_p 1.0, top_k 40 and presence penalty 2.0. Keep these task-specific presets distinct; larger presence penalties can cause language mixing or a slight quality decrease.
Recommended generation allowances are 32,768 tokens for most questions and 81,920 for unusually complex mathematics or programming. These are not separately established hard output ceilings. Define the final-answer form explicitly and check calculations or generated code against the supplied conditions.
Video and substantive limits
The supported video example defaults to two frames per second. The maker’s hour-scale video guidance expands the processor budget to roughly 224K video tokens for higher-rate sampling. That requires configuration and consumes context alongside the task and answer; it is not an unconditional duration or perfect-event-recall guarantee. Preserve relevant scene transitions and small writing, and request a result tied to the visible evidence. The practical value is the checkpoint’s joint text/visual scope and selectable response mode, with a concrete artifact that can be reviewed.
POST /v1/chat/completionstexttext| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen/qwen3.5-35b-a3b | 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.
Match each charge to the conditions shown. Use one rate for each set of alternative configurations.
| Configuration | Rate schedule | Context tokens, more than | Context tokens, up to and including | Charge |
|---|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Context tokens, more than— | Context tokens, up to and includingNo upper bound | Charge Input · Input tokens $0.00046 / 1K tokens Output · Output tokens $0.00366 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
| ConfigurationRequest-length tier 1 | Rate schedule— | Context tokens, more than0 | Context tokens, up to and includingNo upper bound | Charge Input · Input tokens $0.00046 / 1K tokens Output · Output tokens $0.00366 / 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": "qwen/qwen3.5-35b-a3b", "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='qwen/qwen3.5-35b-a3b',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='qwen/qwen3.5-35b-a3b',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 qwen/qwen3.5-35b-a3b
Use the method paths shown above.
qwen/qwen3.5-35b-a3bQwen3.5-35B-A3B is Qwen’s vision-language mixture-of-experts model with 35 billion total and 3 billion activated parameters. It accepts text, images and video for reasoning, coding and tool-assisted work, with thinking and direct-answer modes and a native 262,144-token context.
This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below.
Qwen3.5 highlights: a sparse vision-language checkpoint
Qwen3.5-35B-A3B is a post-trained vision-language model combining text, image and video understanding with textual responses. The family is trained with early text–vision fusion and reinforcement learning across progressively complex tasks, supporting reasoning, coding, agents and visual analysis. It also describes coverage of 201 languages and dialects. That coverage is a family capability rather than a promise of equal quality for every language or task.
The exact open checkpoint is separate from Qwen3.5-35B-A3B-Base and the later Qwen3.6 release. The maker also describes a managed Flash offering with additional features; those interface features should not be assigned automatically to this open checkpoint. Preserve the requested 35B-A3B identity when comparing results or retaining a reproducible model selection.
Model overview
The mixture-of-experts model has 35 billion parameters in total and activates three billion. Its 40-layer hybrid layout repeats three Gated DeltaNet expert layers followed by one gated-attention expert layer. It has 256 routed experts, with eight routed experts and one shared expert active in the documented layout. A vision encoder and multi-token prediction are part of the model. The activated count does not describe the complete checkpoint size, and architecture alone does not guarantee latency in a particular application.
The native context is 262,144 tokens, with conditional extension to 1,010,000. Qwen advises retaining at least a 128K context where possible for complex thinking tasks, even if a smaller window would reduce memory pressure. Text, represented visual material, history, reasoning and the final answer all need space in the configured sequence.
Using inputs and response modes
For a screenshot-informed coding request, provide the visual reference, existing source context, interaction requirements and expected editable artifact. For a chart or document, retain labels and units and request a table of values with supporting evidence. For a multilingual explanation, state the target language and terms to preserve.
Thinking is enabled by default, producing a reasoning block before the final response. The maker’s enable_thinking control can disable it for a direct answer. Qwen3.5 does not officially support the old /think and /nothink soft prompt switches. Its history guidance keeps final answers without thinking content. This should remain distinct from later Qwen3.6 preservation options or Qwen3.8’s default historical reasoning retention.
Agentic usage
The maker documents tool calling and recommends Qwen-Agent with configured or MCP-connected tools, alongside Qwen Code for repository-oriented work. Define the permitted operations, relevant evidence and final deliverable. The application executes tools and supplies their observations; the model can use that feedback to revise its analysis. A planned command or generated patch is not an executed action, and the model name alone does not establish access to current information or files.
Ultra-long context
The maker’s extension recipe uses YaRN with the checkpoint’s interleaved positional settings. A factor of four targets the roughly one-million-token workflow; a factor of two is suggested for 524,288-token tasks. Static scaling can reduce shorter-text performance, so it should be configured when the larger context is needed. Organize long material with document boundaries, code identifiers and a clear question instead of filling the window with unrelated evidence.
Generation recommendations
In Best Practices, general thinking uses temperature 1.0, top_p 0.95, top_k 20, min_p 0, presence penalty 1.5 and repetition penalty 1.0. Precise coding uses temperature 0.6 and presence penalty zero. General direct responses use temperature 0.7, top_p 0.8 and presence penalty 1.5; top_k is 20, min_p is 0 and repetition penalty is 1.0. For reasoning in direct mode, Best Practices give temperature 1.0, top_p 1.0, top_k 40 and presence penalty 2.0. Keep these task-specific presets distinct; larger presence penalties can cause language mixing or a slight quality decrease.
Recommended generation allowances are 32,768 tokens for most questions and 81,920 for unusually complex mathematics or programming. These are not separately established hard output ceilings. Define the final-answer form explicitly and check calculations or generated code against the supplied conditions.
Video and substantive limits
The supported video example defaults to two frames per second. The maker’s hour-scale video guidance expands the processor budget to roughly 224K video tokens for higher-rate sampling. That requires configuration and consumes context alongside the task and answer; it is not an unconditional duration or perfect-event-recall guarantee. Preserve relevant scene transitions and small writing, and request a result tied to the visible evidence. The practical value is the checkpoint’s joint text/visual scope and selectable response mode, with a concrete artifact that can be reviewed.
POST /v1/messagestextjson, sse| Field | Type | Requirement | Values & limits | Default |
|---|---|---|---|---|
Fieldmodel | Typestring | RequirementRequired | Values & limits qwen/qwen3.5-35b-a3b | 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.
Match each charge to the conditions shown. Use one rate for each set of alternative configurations.
| Configuration | Rate schedule | Context tokens, more than | Context tokens, up to and including | Charge |
|---|---|---|---|---|
| ConfigurationBase token rates | Rate scheduleBase rate | Context tokens, more than— | Context tokens, up to and includingNo upper bound | Charge Input · Input tokens $0.00046 / 1K tokens Output · Output tokens $0.00366 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
| ConfigurationRequest-length tier 1 | Rate schedule— | Context tokens, more than0 | Context tokens, up to and includingNo upper bound | Charge Input · Input tokens $0.00046 / 1K tokens Output · Output tokens $0.00366 / 1K tokens Input · Cache read $0 / 1K tokens Input · Cache write $0 / 1K tokens Input · Cache write, 1 hour — / 1K tokens |
Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. How reservation and settlement work →
{\n "model": "qwen/qwen3.5-35b-a3b",\n "max_tokens": 128,\n "messages": [\n {\n "role": "user",\n "content": "Summarize the incident report."\n }\n ],\n "stream": false\n}set -eu\n: "${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': 'qwen/qwen3.5-35b-a3b', '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': 'qwen/qwen3.5-35b-a3b', '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 qwen/qwen3.5-35b-a3b
Use the method paths shown above.
Technical example in English
Audit a transcribed experiment table\n\nUse this transcribed experiment table: baseline, 100 trials, 12 failures; revision A, 80 trials, 8 failures; revision B, 40 trials, 2 failures. Calculate the observed failure rates and compare them. Explain what these counts establish and what would require additional statistical analysis; do not claim that the smallest observed rate proves the best underlying system.A rate table and a qualified comparison separating observed proportions from unperformed statistical inference.Each rate uses its own trial count, and the conclusion does not treat unequal sample sizes as equivalent proof.