Source: https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/ Canonical: https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/ Language: en # DeepSeek: DeepSeek V3.2 Exp [Markdown](https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/index.md) · [JSON](https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/facts.json) ## [DeepSeek: DeepSeek V3.2 Exp](https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/) DeepSeek V3.2 Exp is the sparse-attention edition built on V3.1 Terminus. It processes text for reasoning, coding and agent tasks while exploring more efficient long-context attention, with a 128K producer context and a separate identity from the final V3.2 release. - [**Ailoot Ultra · OpenAI Chat** `POST /v1/chat/completions` `deepseek-v3.2-exp-thinking`](https://ailoot.ai/models/en/deepseek-v3-2-exp-thinking/#method-0) About the model This description covers the model’s capabilities. Available in Ailoot inputs, outputs and options are listed for the selected method below. DeepSeek V3.2 Exp is DeepSeek’s experimental text-model edition built on V3.1 Terminus. Its defining change is DeepSeek Sparse Attention, introduced to study more efficient training and inference for extended text sequences. The task remains text-based: written instructions and context produce a text answer, code or agent-related output. This is the sparse-attention research edition, separate from the final V3.2 model and its later thinking-with-tools format. DSA uses a lightweight indexer and fine-grained token selection so the main attention calculation focuses on selected earlier token information. DeepSeek deliberately aligned the experimental training configuration with V3.1 Terminus to isolate the architectural change. The V3.2 report confirms that the final V3.2 and V3.2 Exp share this architecture and describes continued training from the 128K-context Terminus base. The published experimental profile uses that 128K long-context setting. The producer’s experimental card keeps reasoning and agentic work in scope while preserving the prior model’s task profile. Its evidence covers reasoning without tools and agentic tool use as distinct workflows. Prepare a clear question, relevant source context and the desired final form. If the task uses a tool, the application supplies its definition and returns the result; the model’s proposed action is not the same as an operation that has already executed. The final V3.2 card introduces a revised tool-call format and explicitly describes thinking with tools. Those updates belong to the final release and should not be treated as the experimental edition’s entire interaction contract. For V3.2 Exp, retain the edition’s own message formatting and the inherited Terminus task context rather than assuming that a similar name makes every later control or role interchangeable. For a long document comparison, label each source and identify the sections relevant to the question. An illustrative brief is: “Compare these requirements with the proposed implementation, identify each mismatch and cite the passage that establishes it. End with a prioritised correction list.” For code, provide the failing behaviour, relevant fragment and expected result, then ask for a focused change with a clear explanation. The source layout and completion rule help direct a long-context task toward a usable artifact. The producer’s reference implementation includes the sparse-attention indexer, and its card records an important implementation correction: indexer RoPE uses a non-interleaved layout, whereas the latent-attention module expects an interleaved layout. This is a producer-reference detail relevant to faithful model execution, rather than a different user prompt or a new capability. It also means a model name alone does not describe every implementation’s handling of the experimental architecture. A continuing agent task must budget its source text, conversation, generated reasoning and tool observations together. Keep the useful observations organised and ask for the final answer to distinguish evidence, assumptions and proposed actions. For a source-dependent factual result, supply the authoritative material or a suitable tool rather than relying on broad training knowledge. The output remains a text, code or analytical record that can be checked against the task and its inputs. **Ailoot Ultra · OpenAI Chat** `POST /v1/chat/completions` `deepseek-v3.2-exp-thinking` [`deepseek-v3.2-exp-thinking`](https://ailoot.ai/models/en/#m=deepseek-v3.2-exp-thinking&route=Ultra&caller=deepseek-v3.2-exp-thinking&fmt=chat) DeepSeek /`deepseek-v3.2-exp-thinking` ### Overview DeepSeek V3.2 Exp is the sparse-attention edition built on V3.1 Terminus. It processes text for reasoning, coding and agent tasks while exploring more efficient long-context attention, with a 128K producer context and a separate identity from the final V3.2 release. Accepted inputs Text Results Text ### Available in Ailoot · inputs & methods **Ailoot Ultra · OpenAI Chat** `POST /v1/chat/completions` Input types `text` Output types `text` Output format json or SSE | Field | Type | Requirement | Values & limits | Default | | --- | --- | --- | --- | --- | | Field`model` | Type`string` | RequirementRequired | Values & limits deepseek-v3.2-exp-thinking | Default— | | Field`messages` | Type`message[]` | RequirementRequired | Values & limits Minimum items: 1 | Default— | | Field`messages[].role` | Type`string` | RequirementRequired | Values & limits Model-specific | Default— | | Field`messages[].content` | Type`string` | RequirementOptional | Values & limits Applies when: Text conversation | Default— | | Field`stream` | Type`boolean` | RequirementOptional | Values & limits | Defaultfalse | Input material combinations **Text** · UTF8 text `messages[].content` Method parameters and lifecycle | Parameter | Contract | | --- | --- | | Parameter`content_type` | Contractapplication/json | | Parameter`required` | Contractmodel, messages | | Parameter`material_types` | Contracttext | | Parameter`messages` | Contractmessage array with role/content shape | | Parameter`file_material` | Contractactual UTF-8 document inserted as literal text; no file ID inferred | | Parameter`optional_controls` | Contractstream | | Parameter`control_scope` | ContractText conversation; stream is a boolean and defaults to false. | | Parameter`output_modalities` | Contracttext | | Parameter`hosted_operations` | Contractfalse | | Parameter`multimodal_input_binding` | Contractnull | Method aliases: `/chat/completions` 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. ### Pricing Tariff references do not add request options. Supported parameters are listed for the selected method. Fast mode Not specified When fast mode is selected Flex mode Not specified When flex mode is selected Maximum reasoning effort Not specified When maximum reasoning effort uses the default adjustment #### Ailoot rates #### Prices by configuration | Configuration | Rate schedule | Charge | | --- | --- | --- | | ConfigurationBase token rates | Rate scheduleBase rate | Charge Input · Input tokens **$0.00058** / 1K tokens Output · Output tokens **$0.00086** / 1K tokens Input · Cache read **$0** / 1K tokens Input · Cache write **$0** / 1K tokens Input · Cache write, 1 hour **—** / 1K tokens | Ailoot reserves an estimated amount before a request. After the result, it settles the final charge from measured usage. Streaming settlement uses the final usage record. [How reservation and settlement work →](https://ailoot.ai/docs/en/billing/reserve/) ### Example requests #### Ailoot Ultra · OpenAI Chat curlPython SDK ```bash set -eu : "${AILOOT_API_KEY:?Set your Ailoot API key}" python3 - <<'PY_REQUEST' import json,os from pathlib import Path p=json.loads('{"model":"deepseek-v3.2-exp-thinking","messages":[{"role":"user","content":"hello"}],"stream":false}') Path("request.json").write_text(json.dumps(p,ensure_ascii=False)) PY_REQUEST curl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/chat/completions \ -H "Authorization: Bearer $AILOOT_API_KEY" -H 'Content-Type: application/json' \ --data-binary @request.json -D response-headers.txt -o result.json python3 - <<'PY_RESULT' import json from pathlib import Path r=json.loads(Path("result.json").read_text()) for x in r.get("choices",[]): if x.get("message",{}).get("content"): print(x["message"]["content"]) PY_RESULT ``` ```python from openai import OpenAI import os if not os.environ.get("AILOOT_API_KEY", "").strip(): raise SystemExit("Set AILOOT_API_KEY before making a request") client=OpenAI(base_url="https://api.ailoot.ai/v1",api_key=os.environ["AILOOT_API_KEY"],max_retries=0) result=client.chat.completions.create(model="deepseek-v3.2-exp-thinking", messages=[{"role":"user","content":"hello"}], stream=False) import base64, json from pathlib import Path def save_result(data, stem="result"): # Keep every field, including usage, IDs, reasoning and tool requests. Path(stem + ".json").write_text(json.dumps(data, ensure_ascii=False, indent=2), encoding="utf-8") print(json.dumps(data, ensure_ascii=False, indent=2)) def save_blob(encoded, suffix, label): raw = base64.b64decode(encoded, altchars=b"-_", validate=True) 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 file = Path(stem + "-" + label + "." + ext) file.write_bytes(raw) print("Saved", file) for index, block in enumerate(data.get("content") or []): print("Content", index, json.dumps(block, ensure_ascii=False)) for choice in data.get("choices") or []: print("Choice", choice.get("index"), json.dumps(choice.get("message", choice.get("delta", {})), ensure_ascii=False)) for index, item in enumerate(data.get("output") or []): print("Output", index, json.dumps(item, ensure_ascii=False)) if item.get("type") == "image_generation_call" and item.get("result"): save_blob(item["result"], item.get("output_format") or "bin", "output-" + str(index)) for ci, candidate in enumerate(data.get("candidates") or []): for pi, part in enumerate((candidate.get("content") or {}).get("parts") or []): print("Candidate/part", ci, pi, json.dumps(part, ensure_ascii=False)) blob = part.get("inlineData") or part.get("inline_data") if blob and blob.get("data"): mime = blob.get("mimeType") or blob.get("mime_type") suffix = {"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"}.get(mime, "bin") save_blob(blob["data"], suffix, "candidate-" + str(ci) + "-part-" + str(pi)) save_result(result.to_dict(mode="json")) ``` ### Receive and save results #### Ailoot Ultra · OpenAI Chat json or SSE ##### Keep all completed results 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. ##### If the original response needs checking Keep request/response IDs and headers. No generic text job retrieval URL is supplied. Response fields | Field | Type | | --- | --- | | Field`choices[].index` | Typeinteger | | Field`choices[].message.content` | Typestring\|null | | Field`choices[].message.tool_calls` | Typetool\_call\[\] | | Field`choices[].message.reasoning_content` | Typestring | | Field`choices[].message.reasoning_details` | Typeobject\|array | | Field`choices[].finish_reason` | Typestring\|null | | Field`usage` | Typeobject | Errors and next steps | 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. | ### Aliases #### Model aliases **Ailoot Ultra** [`deepseek-v3.2-exp-thinking`](https://ailoot.ai/models/en/#m=deepseek-v3.2-exp-thinking&route=Ultra&caller=deepseek-v3.2-exp-thinking&fmt=chat) #### Method path aliases `/chat/completions` #### Related caller IDs [`deepseek-v3.2-exp-thinking`](https://ailoot.ai/models/en/#m=deepseek-v3.2-exp-thinking&route=Ultra&caller=deepseek-v3.2-exp-thinking&fmt=chat) FAQ · DeepSeek: DeepSeek V3.2 Exp What inputs and results are available? `POST /v1/chat/completions` · `deepseek-v3.2-exp-thinking`: `text` → `text` Which fields does the request use? `POST /v1/chat/completions` · `deepseek-v3.2-exp-thinking`: `model`, `messages`, `stream` What limits apply? `POST /v1/chat/completions` · `deepseek-v3.2-exp-thinking`: `messages.min_items`: `1` Connect Ailoot to Claude Code, OpenCode, Hermes or your own app. ## The company and its project portfolio The company and its project portfolio: millions of customers; monthly sales in the billions. ## Buy tokens at a discount. Subscription discounts for regular use. Promo codes offer up to 50% off token purchases. Subscription Choose a discounted plan for your regular requests. Promo code: up to 50% off Enter your code at checkout. The price updates to reflect the current promotion. Spending in your history The model, tokens and amount are recorded for each request. [Choose a subscription](https://ailoot.ai/docs/en/billing/subscriptions/) · [View model prices](https://ailoot.ai/models/en/) ## Connect the API to a product you sell. Use Ailoot in SaaS, apps and bots under its official service offer and licence agreement. Request a reseller licence for resale. Commercial application Add text, image, video and audio generation to paid features in your service. Reseller licence Send a description of your project and how you provide access. Request written terms for use and resale. Documents on the website The service offer, licence agreement and data processing policy are available before you connect. [Request a reseller licence](mailto:support@ailoot.ai) · [Service offer and licence agreement](https://ailoot.ai/legal/offer?lang=en) ## Ailoot · The company and its project portfolio Ailoot combines established operating experience, a large client base, stable service and ongoing KPI control. The company and its project portfolio: millions of customers; monthly sales in the billions. Ailoot brings that operating experience to API access for AI features in customer products, agents, applications and automations.