DeepSeek V4.1 Flash — boshidan multimodal model: vizual enkoder tasvirni matn bilan birgalikda qayta ishlanadigan ifodaga aylantiradi. Natija matnli. Vizual tushunish, dasturlash va agent vazifalari ko‘zda tutilgan; kontekst bir million tokengacha yetadi.
Asosiy tarmoq 552 mlrd parametrni o‘z ichiga oladi. Sababiy enkoder va dekoderning har biri 20 ta qatlamga ega. Kirishni qayta ishlashda har bir token uchun 8 mlrd, yaratishda esa 16 mlrd parametr faol bo‘ladi. Bunday ajratish kiruvchi materiallari javobdan ancha katta bo‘lgan vazifalar uchun mo‘ljallangan.
Siqilgan siyrak diqqat mexanizmi qatlamlar orasidagi ma’lumotlarni qayta ishlatadi va kontekstning muhim qismlarini takroriy qidirishni cheklaydi. Mahalliy diqqatning yetishmayotgan holatlari yaqindagi parchadan tiklanadi. Engram shartli xotirasi 196 mlrd parametrni o‘z ichiga oladi va tokenlar bo‘yicha siyrak qidiruv orqali yozuvlarga murojaat qiladi. Bu asosiy tarmoq hajmining yangi nomi emas, alohida xotira blokidir. Har bir ekspert qatlamida bitta umumiy va 384 ta marshrutlanadigan ekspert bor; token uchun oltita marshrutlanadigan ekspert tanlanadi. Bundan tashqari, qatlamlar orasidagi o‘zgartirilgan bog‘lanishlar va spekulyativ dekodlash uchun DSpark ishlatiladi.
Model 45 trln tokenli multimodal korpusda noldan o‘qitilgan. Keyingi tayyorlash tartibi: misollar orqali o‘qitish, mustahkamlash orqali o‘qitish, so‘ng modelning o‘z ketma-ketliklarida distillyatsiya.
DeepSeek sintetik vazifa ma’lumotlari, ish muhitlari va agent yo‘llarini kengaytiradi. Tayyorlashda topshiriqning o‘zi, mavjud muhit va natijani tekshirish usuli bog‘lanadi; ma’lumotlar saralanadi, takrorlardan tozalanadi va murakkabligiga qarab tanlanadi. Bu modelga murojaat qilganda tashqi vositalarning avtomatik mavjudligi emas, o‘qitish tuzilishidir.
Mulohaza yuritish 1 dan 100 gacha butun son bilan boshqariladi. Bu uzluksiz shkala tahlil uchun sa’y-harakatni belgilaydi; chuqur mulohaza ko‘proq vaqt va token talab qilishi mumkin.
Model formati ko‘p navbatli suhbat, vosita chaqiruvlari, saqlangan mulohaza va matn bilan tasvirlarning almashib kelishini ko‘zda tutadi. Vositalarni ilova bajaradi; olingan natijalar keyingi qadam uchun materialga aylanadi.
Javob matnli bo‘lib qoladi. Vizual qobiliyat tegishli kirish metodini talab qiladi. Kontekst hajmi va tavsiya etilgan yaratish uzunligi javobning universal maksimal hajmini belgilamaydi.
", "aboutText": "### Matn va tasvirlarni birgalikda qayta ishlash\n\nDeepSeek V4.1 Flash — boshidan multimodal model: vizual enkoder tasvirni matn bilan birgalikda qayta ishlanadigan ifodaga aylantiradi. Natija matnli. Vizual tushunish, dasturlash va agent vazifalari ko‘zda tutilgan; kontekst bir million tokengacha yetadi.\n\n### Katta kirish uchun arxitektura\n\nAsosiy tarmoq 552 mlrd parametrni o‘z ichiga oladi. Sababiy enkoder va dekoderning har biri 20 ta qatlamga ega. Kirishni qayta ishlashda har bir token uchun 8 mlrd, yaratishda esa 16 mlrd parametr faol bo‘ladi. Bunday ajratish kiruvchi materiallari javobdan ancha katta bo‘lgan vazifalar uchun mo‘ljallangan.\n\nSiqilgan siyrak diqqat mexanizmi qatlamlar orasidagi ma’lumotlarni qayta ishlatadi va kontekstning muhim qismlarini takroriy qidirishni cheklaydi. Mahalliy diqqatning yetishmayotgan holatlari yaqindagi parchadan tiklanadi. Engram shartli xotirasi 196 mlrd parametrni o‘z ichiga oladi va tokenlar bo‘yicha siyrak qidiruv orqali yozuvlarga murojaat qiladi. Bu asosiy tarmoq hajmining yangi nomi emas, alohida xotira blokidir. Har bir ekspert qatlamida bitta umumiy va 384 ta marshrutlanadigan ekspert bor; token uchun oltita marshrutlanadigan ekspert tanlanadi. Bundan tashqari, qatlamlar orasidagi o‘zgartirilgan bog‘lanishlar va spekulyativ dekodlash uchun DSpark ishlatiladi.\n\n### O‘qitish va agent muhitlari\n\nModel 45 trln tokenli multimodal korpusda noldan o‘qitilgan. Keyingi tayyorlash tartibi: misollar orqali o‘qitish, mustahkamlash orqali o‘qitish, so‘ng modelning o‘z ketma-ketliklarida distillyatsiya.\n\nDeepSeek sintetik vazifa ma’lumotlari, ish muhitlari va agent yo‘llarini kengaytiradi. Tayyorlashda topshiriqning o‘zi, mavjud muhit va natijani tekshirish usuli bog‘lanadi; ma’lumotlar saralanadi, takrorlardan tozalanadi va murakkabligiga qarab tanlanadi. Bu modelga murojaat qilganda tashqi vositalarning avtomatik mavjudligi emas, o‘qitish tuzilishidir.\n\n### Tahlilning boshqariladigan chuqurligi\n\nMulohaza yuritish 1 dan 100 gacha butun son bilan boshqariladi. Bu uzluksiz shkala tahlil uchun sa’y-harakatni belgilaydi; chuqur mulohaza ko‘proq vaqt va token talab qilishi mumkin.\n\n### Suhbatni davom ettirish va amallar\n\nModel formati ko‘p navbatli suhbat, vosita chaqiruvlari, saqlangan mulohaza va matn bilan tasvirlarning almashib kelishini ko‘zda tutadi. Vositalarni ilova bajaradi; olingan natijalar keyingi qadam uchun materialga aylanadi.\n\nJavob matnli bo‘lib qoladi. Vizual qobiliyat tegishli kirish metodini talab qiladi. Kontekst hajmi va tavsiya etilgan yaratish uzunligi javobning universal maksimal hajmini belgilamaydi.", "editorialOverlay": { "copy_fields": { "practical_html": "Ingliz tilidagi texnik misol
Trace a data-loss bug from a text log\n\nUse this log and code sketch to identify the likely lost-update mechanism: A reads stock=5; B reads stock=5; A writes 4; B writes 4. Both requests call read(), subtract one, then write(). Explain the failure, propose one atomic update strategy and a concurrent test. Do not assume a specific database or execute anything.A causal diagnosis, database-neutral pseudocode for an atomic decrement and a concurrent-test scenario with the expected final count.The test expects stock 3 after two successful decrements from 5, with a stated insufficient-stock rule.deepseek-ai/deepseek-v4.1-flashDeepSeek V4.1 Flash — matn va tasvirlarni tahlil qilish, dasturlash va uzun kontekstli agent ishi uchun multimodal model.
Bu tavsif model imkoniyatlarini yoritadi. Ailoot da tanlangan usul uchun mavjud kirish, chiqish va parametrlar quyida keltirilgan.
DeepSeek V4.1 Flash — boshidan multimodal model: vizual enkoder tasvirni matn bilan birgalikda qayta ishlanadigan ifodaga aylantiradi. Natija matnli. Vizual tushunish, dasturlash va agent vazifalari ko‘zda tutilgan; kontekst bir million tokengacha yetadi.
Asosiy tarmoq 552 mlrd parametrni o‘z ichiga oladi. Sababiy enkoder va dekoderning har biri 20 ta qatlamga ega. Kirishni qayta ishlashda har bir token uchun 8 mlrd, yaratishda esa 16 mlrd parametr faol bo‘ladi. Bunday ajratish kiruvchi materiallari javobdan ancha katta bo‘lgan vazifalar uchun mo‘ljallangan.
Siqilgan siyrak diqqat mexanizmi qatlamlar orasidagi ma’lumotlarni qayta ishlatadi va kontekstning muhim qismlarini takroriy qidirishni cheklaydi. Mahalliy diqqatning yetishmayotgan holatlari yaqindagi parchadan tiklanadi. Engram shartli xotirasi 196 mlrd parametrni o‘z ichiga oladi va tokenlar bo‘yicha siyrak qidiruv orqali yozuvlarga murojaat qiladi. Bu asosiy tarmoq hajmining yangi nomi emas, alohida xotira blokidir. Har bir ekspert qatlamida bitta umumiy va 384 ta marshrutlanadigan ekspert bor; token uchun oltita marshrutlanadigan ekspert tanlanadi. Bundan tashqari, qatlamlar orasidagi o‘zgartirilgan bog‘lanishlar va spekulyativ dekodlash uchun DSpark ishlatiladi.
Model 45 trln tokenli multimodal korpusda noldan o‘qitilgan. Keyingi tayyorlash tartibi: misollar orqali o‘qitish, mustahkamlash orqali o‘qitish, so‘ng modelning o‘z ketma-ketliklarida distillyatsiya.
DeepSeek sintetik vazifa ma’lumotlari, ish muhitlari va agent yo‘llarini kengaytiradi. Tayyorlashda topshiriqning o‘zi, mavjud muhit va natijani tekshirish usuli bog‘lanadi; ma’lumotlar saralanadi, takrorlardan tozalanadi va murakkabligiga qarab tanlanadi. Bu modelga murojaat qilganda tashqi vositalarning avtomatik mavjudligi emas, o‘qitish tuzilishidir.
Mulohaza yuritish 1 dan 100 gacha butun son bilan boshqariladi. Bu uzluksiz shkala tahlil uchun sa’y-harakatni belgilaydi; chuqur mulohaza ko‘proq vaqt va token talab qilishi mumkin.
Model formati ko‘p navbatli suhbat, vosita chaqiruvlari, saqlangan mulohaza va matn bilan tasvirlarning almashib kelishini ko‘zda tutadi. Vositalarni ilova bajaradi; olingan natijalar keyingi qadam uchun materialga aylanadi.
Javob matnli bo‘lib qoladi. Vizual qobiliyat tegishli kirish metodini talab qiladi. Kontekst hajmi va tavsiya etilgan yaratish uzunligi javobning universal maksimal hajmini belgilamaydi.
POST /v1/messagestexttext| Maydon | Tur | Talab | Qiymatlar va cheklovlar | Standart |
|---|---|---|---|---|
Maydonmodel | Turstring | TalabMajburiy | Qiymatlar va cheklovlar deepseek-ai/deepseek-v4.1-flash | Standart— |
Maydonmessages | Turmessage[] | TalabMajburiy | Qiymatlar va cheklovlar Elementlarning eng kam soni: 1 | Standart— |
Maydonmessages[].role | Turstring | TalabMajburiy | Qiymatlar va cheklovlar user, assistant | Standart— |
Maydonmessages[].content | Turstring|content_block[] | TalabMajburiy | Qiymatlar va cheklovlar | Standart— |
Maydonsystem | Turstring|content_block[] | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydonmax_tokens | Turinteger | TalabMajburiy | Qiymatlar va cheklovlar Eng kam: 1 Cheklov qo‘llanadigan obyekt: Tanlangan model yoki kartaning aniq yuqori chegarasi | Standart— |
Maydonstream | Turboolean | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standartfalse |
Maydontemperature | Turnumber | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontop_p | Turnumber | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonstop_sequences | Turstring[] | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontools | Turtool[] | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontools[].name | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Quyidagi holatda qo‘llanadi: Vosita taqdim etilgan | Standart— |
Maydontools[].description | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydontools[].input_schema | TurJSON_schema | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydontool_choice | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonthinking.type | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonthinking.budget_tokens | Turinteger | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonoutput_config.effort | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonmetadata | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydoncache_control | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Matn · UTF8 matni
messages[].content yoki type=text bloki
| Parametr | Parametr talablari |
|---|---|
Parametrmaterial_types | Parametr talablaritext |
Parametrrequired | Parametr talablarimodel, messages, max_tokens |
Parametroptional_controls | Parametr talablarinull |
Parametrcontent_type | Parametr talablariapplication/json |
Parametrtext_file_material | Parametr talablariHaqiqiy UTF-8 kontenti matn sifatida kiritiladi; fayl IDlari/URLs uchun alohida aniq qo‘llab-quvvatlash kerak |
Hayot sikli: exact retained published method
Metod tuzilishlari va IDlari aniq; yetkazib beruvchining imkoniyati so‘rovni qabul qilishga ruxsatdan alohida.
POST /v1/messagesAiloot so‘rovdan oldin taxminiy summani band qiladi. Natijadan keyin o‘lchangan sarf asosidagi yakuniy to‘lovni yechadi. Oqimli javob bo‘yicha yakuniy hisob-kitob yakuniy sarf yozuviga asoslanadi. Summani band qilish va to‘lovni hisoblash tartibi →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 source-material file}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "deepseek-ai/deepseek-v4.1-flash", "max_tokens": 1024, "messages": []}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["messages"]=[{"role":"user","content":"Summarize this material and quote supporting source passages.\\n\\n"+text}]\nPath("request.json").write_text(json.dumps(p,ensure_ascii=False))\nPY_REQUEST\ncurl --fail-with-body --silent --show-error https://api.ailoot.ai/v1/messages \\\n -H "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\nprint(json.dumps(json.loads(Path("result.json").read_text()),ensure_ascii=False,indent=2))\nPY_RESULT\nimport anthropic\nimport os\n\nif not os.environ.get("AILOOT_API_KEY", "").strip(): raise SystemExit("Set AILOOT_API_KEY before making a request")\nclient = anthropic.Anthropic(base_url="https://api.ailoot.ai", api_key=os.environ["AILOOT_API_KEY"], max_retries=0)\nmsg = client.messages.create(\n model="deepseek-ai/deepseek-v4.1-flash",\n max_tokens=1024,\n messages=[{"role": "user", "content": "Explain the difference between an observation and a hypothesis in three sentences."}],\n)\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(msg.to_dict(mode="json"))\nimport Anthropic from "@anthropic-ai/sdk";\n\nif (!process.env.AILOOT_API_KEY?.trim()) throw new Error("Set AILOOT_API_KEY before making a request");\nconst client = new Anthropic({ baseURL: "https://api.ailoot.ai", apiKey: process.env.AILOOT_API_KEY, maxRetries: 0 });\nconst msg = await client.messages.create({\n model: "deepseek-ai/deepseek-v4.1-flash",\n max_tokens: 1024,\n messages: [{ role: "user", content: "Explain the difference between an observation and a hypothesis in three sentences." }],\n});\nimport { writeFileSync } from "node:fs";\n\nfunction saveResult(data, stem = "result") {\n writeFileSync(stem + ".json", JSON.stringify(data, null, 2));\n console.log(JSON.stringify(data, null, 2));\n const saveBlob = (encoded, suffix, label) => {\n const raw = Buffer.from(encoded, "base64");\n const ext = raw.subarray(0,8).equals(Buffer.from([137,80,78,71,13,10,26,10])) ? "png" : raw[0]===255 && raw[1]===216 ? "jpg" : raw.subarray(0,4).toString()==="RIFF" && raw.subarray(8,12).toString()==="WEBP" ? "webp" : suffix;\n const file = `${stem}-${label}.${ext}`;\n writeFileSync(file, raw); console.log("Saved", file);\n };\n for (const [i, block] of (data.content || []).entries()) console.log("Content", i, JSON.stringify(block));\n for (const choice of data.choices || []) console.log("Choice", choice.index, JSON.stringify(choice.message || choice.delta || {}));\n for (const [i, item] of (data.output || []).entries()) {\n console.log("Output", i, JSON.stringify(item));\n if (item.type === "image_generation_call" && item.result) saveBlob(item.result, item.output_format || "bin", `output-${i}`);\n }\n for (const [ci, candidate] of (data.candidates || []).entries()) {\n for (const [pi, part] of (candidate.content?.parts || []).entries()) {\n console.log("Candidate/part", ci, pi, JSON.stringify(part));\n const blob = part.inlineData || part.inline_data;\n if (blob?.data) {\n const mime = blob.mimeType || blob.mime_type;\n const suffix = ({"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"})[mime] || "bin";\n saveBlob(blob.data, suffix, `candidate-${ci}-part-${pi}`);\n }\n }\n }\n}\n\nsaveResult(msg);\nimport anthropic,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=anthropic.Anthropic(base_url="https://api.ailoot.ai",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.messages.stream(model='deepseek-ai/deepseek-v4.1-flash',max_tokens=1024,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_message()\n save_result(final.to_dict(mode="json"),"stream-result")\njson yoki SSE
Qaytarilgan har bir to‘liq javob, fayl va kontent qismini saqlang.
Barcha kontent bloklari, fikrlash imzolari va haqiqiy vosita ishlatish IDlarini saqlang. SSE bloklarining barcha o‘zgarish qismlarini yig‘ing; birinchi matn bloki to‘liq chiqish emas.
So‘rov va javob IDlari hamda sarlavhalarini saqlang. Matn vazifasini olish uchun umumiy URL taqdim etilmagan.
| Maydon | Tur |
|---|---|
Maydonid | Turstring |
Maydonmodel | Turstring |
Maydoncontent[] | Turtext|thinking|tool_use|other content block |
Maydonstop_reason | Turstring|null |
Maydonusage | Turobject |
| Javob | Shart | Keyingi qadam |
|---|---|---|
Javob400 · invalid_request_error | ShartMajburiy maydonlar, material ifodasi yoki parametr qiymati noto‘g‘ri | Keyingi qadamYangi yaratish so‘rovidan oldin ko‘rsatilgan maydonlarni tuzating. |
Ailoot deepseek-ai/deepseek-v4.1-flash
Yuqorida ko‘rsatilgan usul yo‘llaridan foydalaning.
deepseek-ai/deepseek-v4.1-flashDeepSeek V4.1 Flash — matn va tasvirlarni tahlil qilish, dasturlash va uzun kontekstli agent ishi uchun multimodal model.
Bu tavsif model imkoniyatlarini yoritadi. Ailoot da tanlangan usul uchun mavjud kirish, chiqish va parametrlar quyida keltirilgan.
DeepSeek V4.1 Flash — boshidan multimodal model: vizual enkoder tasvirni matn bilan birgalikda qayta ishlanadigan ifodaga aylantiradi. Natija matnli. Vizual tushunish, dasturlash va agent vazifalari ko‘zda tutilgan; kontekst bir million tokengacha yetadi.
Asosiy tarmoq 552 mlrd parametrni o‘z ichiga oladi. Sababiy enkoder va dekoderning har biri 20 ta qatlamga ega. Kirishni qayta ishlashda har bir token uchun 8 mlrd, yaratishda esa 16 mlrd parametr faol bo‘ladi. Bunday ajratish kiruvchi materiallari javobdan ancha katta bo‘lgan vazifalar uchun mo‘ljallangan.
Siqilgan siyrak diqqat mexanizmi qatlamlar orasidagi ma’lumotlarni qayta ishlatadi va kontekstning muhim qismlarini takroriy qidirishni cheklaydi. Mahalliy diqqatning yetishmayotgan holatlari yaqindagi parchadan tiklanadi. Engram shartli xotirasi 196 mlrd parametrni o‘z ichiga oladi va tokenlar bo‘yicha siyrak qidiruv orqali yozuvlarga murojaat qiladi. Bu asosiy tarmoq hajmining yangi nomi emas, alohida xotira blokidir. Har bir ekspert qatlamida bitta umumiy va 384 ta marshrutlanadigan ekspert bor; token uchun oltita marshrutlanadigan ekspert tanlanadi. Bundan tashqari, qatlamlar orasidagi o‘zgartirilgan bog‘lanishlar va spekulyativ dekodlash uchun DSpark ishlatiladi.
Model 45 trln tokenli multimodal korpusda noldan o‘qitilgan. Keyingi tayyorlash tartibi: misollar orqali o‘qitish, mustahkamlash orqali o‘qitish, so‘ng modelning o‘z ketma-ketliklarida distillyatsiya.
DeepSeek sintetik vazifa ma’lumotlari, ish muhitlari va agent yo‘llarini kengaytiradi. Tayyorlashda topshiriqning o‘zi, mavjud muhit va natijani tekshirish usuli bog‘lanadi; ma’lumotlar saralanadi, takrorlardan tozalanadi va murakkabligiga qarab tanlanadi. Bu modelga murojaat qilganda tashqi vositalarning avtomatik mavjudligi emas, o‘qitish tuzilishidir.
Mulohaza yuritish 1 dan 100 gacha butun son bilan boshqariladi. Bu uzluksiz shkala tahlil uchun sa’y-harakatni belgilaydi; chuqur mulohaza ko‘proq vaqt va token talab qilishi mumkin.
Model formati ko‘p navbatli suhbat, vosita chaqiruvlari, saqlangan mulohaza va matn bilan tasvirlarning almashib kelishini ko‘zda tutadi. Vositalarni ilova bajaradi; olingan natijalar keyingi qadam uchun materialga aylanadi.
Javob matnli bo‘lib qoladi. Vizual qobiliyat tegishli kirish metodini talab qiladi. Kontekst hajmi va tavsiya etilgan yaratish uzunligi javobning universal maksimal hajmini belgilamaydi.
POST /v1/chat/completionstexttext| Maydon | Tur | Talab | Qiymatlar va cheklovlar | Standart |
|---|---|---|---|---|
Maydonmodel | Turstring | TalabMajburiy | Qiymatlar va cheklovlar deepseek-ai/deepseek-v4.1-flash | Standart— |
Maydonmessages | Turmessage[] | TalabMajburiy | Qiymatlar va cheklovlar Elementlarning eng kam soni: 1 | Standart— |
Maydonmessages[].role | Turstring | TalabMajburiy | Qiymatlar va cheklovlar Namunalar: system, user, assistant, tool, function Shartlar: Bu rollar manba kodining moslik turlarida ochiq ifodalangan; aniq rollar to‘plami har bir modelga xos bo‘lib qoladi. Modelga bogʻliq | Standart— |
Maydonmessages[].content | Turstring|content_part[]|null | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydonmessages[].tool_calls | Turtool_call[] | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydonmessages[].tool_call_id | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Quyidagi holatda qo‘llanadi: role=tool; haqiqatan qaytarilgan IDdan foydalaning | Standart— |
Maydonmessages[].name | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydonmessages[].reasoning_content | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Quyidagi holatda qo‘llanadi: Ushbu model foydalanganda, haqiqatan qaytarilgan fikrlash ma’lumotlarini keyingi so‘rovda qayta yuboring | Standart— |
Maydonmessages[].reasoning_details | Turobject|array | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonmax_tokens | Turinteger | TalabIxtiyoriy | Qiymatlar va cheklovlar Eng kam: 1 Modelga bogʻliq | Standart— |
Maydonmax_completion_tokens | Turinteger | TalabIxtiyoriy | Qiymatlar va cheklovlar Eng kam: 1 Modelga bogʻliq | Standart— |
Maydonstream | Turboolean | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standartfalse |
Maydonstream_options.include_usage | Turboolean | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontemperature | Turnumber | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontop_p | Turnumber | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonstop | Turstring|string[] | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontools | Turfunction_tool[] | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontools[].type | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydontools[].function.name | Turstring | TalabMajburiy | Qiymatlar va cheklovlar Quyidagi holatda qo‘llanadi: Funksiya vositasi taqdim etilgan | Standart— |
Maydontools[].function.description | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydontools[].function.parameters | TurJSON_schema | TalabIxtiyoriy | Qiymatlar va cheklovlar | Standart— |
Maydontool_choice | Turstring|object | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonparallel_tool_calls | Turboolean | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonresponse_format | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonreasoning_effort | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonreasoning | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonservice_tier | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydoninstructions | Turstring | TalabIxtiyoriy | Qiymatlar va cheklovlar Moslik maydoni | Standart— |
Maydonfunctions | Turfunction[] | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonfunction_call | Turstring|object | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Maydonprompt_cache_options | Turobject | TalabIxtiyoriy | Qiymatlar va cheklovlar Modelga bogʻliq | Standart— |
Matn · UTF8 matni
messages[].content
| Parametr | Parametr talablari |
|---|---|
Parametrmaterial_types | Parametr talablaritext |
Parametrrequired | Parametr talablarimodel, messages |
Parametroptional_controls | Parametr talablarinull |
Parametrcontent_type | Parametr talablariapplication/json |
Parametrtext_file_material | Parametr talablariHaqiqiy UTF-8 kontenti matn sifatida kiritiladi; fayl IDlari/URLs uchun alohida aniq qo‘llab-quvvatlash kerak |
Hayot sikli: exact retained published method
Metod tuzilishlari va IDlari aniq; yetkazib beruvchining imkoniyati so‘rovni qabul qilishga ruxsatdan alohida.
POST /v1/chat/completionsAiloot so‘rovdan oldin taxminiy summani band qiladi. Natijadan keyin o‘lchangan sarf asosidagi yakuniy to‘lovni yechadi. Oqimli javob bo‘yicha yakuniy hisob-kitob yakuniy sarf yozuviga asoslanadi. Summani band qilish va to‘lovni hisoblash tartibi →
set -eu\n: "${AILOOT_API_KEY:?Set your Ailoot API key}"\n: "${AILOOT_TEXT_FILE:?Set your actual UTF-8 source-material file}"\nexport AILOOT_TEXT_FILE\npython3 - <<'PY_REQUEST'\nimport json,os\nfrom pathlib import Path\np=json.loads('{"model": "deepseek-ai/deepseek-v4.1-flash", "messages": []}')\nf=Path(os.environ["AILOOT_TEXT_FILE"])\nif not f.is_file():raise SystemExit("Choose your actual UTF-8 source material")\ntext=f.read_text(encoding="utf-8")\nif not text.strip():raise SystemExit("Source material must contain text")\np["messages"]=[{"role":"user","content":"Summarize this material and quote supporting source passages.\\n\\n"+text}]\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\nprint(json.dumps(json.loads(Path("result.json").read_text()),ensure_ascii=False,indent=2))\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)\nresp = client.chat.completions.create(\n model="deepseek-ai/deepseek-v4.1-flash",\n messages=[{"role": "user", "content": "Explain the difference between an observation and a hypothesis in three sentences."}],\n)\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(resp.to_dict(mode="json"))\nimport OpenAI from "openai";\n\nif (!process.env.AILOOT_API_KEY?.trim()) throw new Error("Set AILOOT_API_KEY before making a request");\nconst client = new OpenAI({ baseURL: "https://api.ailoot.ai/v1", apiKey: process.env.AILOOT_API_KEY, maxRetries: 0 });\nconst resp = await client.chat.completions.create({\n model: "deepseek-ai/deepseek-v4.1-flash",\n messages: [{ role: "user", content: "Explain the difference between an observation and a hypothesis in three sentences." }],\n});\nimport { writeFileSync } from "node:fs";\n\nfunction saveResult(data, stem = "result") {\n writeFileSync(stem + ".json", JSON.stringify(data, null, 2));\n console.log(JSON.stringify(data, null, 2));\n const saveBlob = (encoded, suffix, label) => {\n const raw = Buffer.from(encoded, "base64");\n const ext = raw.subarray(0,8).equals(Buffer.from([137,80,78,71,13,10,26,10])) ? "png" : raw[0]===255 && raw[1]===216 ? "jpg" : raw.subarray(0,4).toString()==="RIFF" && raw.subarray(8,12).toString()==="WEBP" ? "webp" : suffix;\n const file = `${stem}-${label}.${ext}`;\n writeFileSync(file, raw); console.log("Saved", file);\n };\n for (const [i, block] of (data.content || []).entries()) console.log("Content", i, JSON.stringify(block));\n for (const choice of data.choices || []) console.log("Choice", choice.index, JSON.stringify(choice.message || choice.delta || {}));\n for (const [i, item] of (data.output || []).entries()) {\n console.log("Output", i, JSON.stringify(item));\n if (item.type === "image_generation_call" && item.result) saveBlob(item.result, item.output_format || "bin", `output-${i}`);\n }\n for (const [ci, candidate] of (data.candidates || []).entries()) {\n for (const [pi, part] of (candidate.content?.parts || []).entries()) {\n console.log("Candidate/part", ci, pi, JSON.stringify(part));\n const blob = part.inlineData || part.inline_data;\n if (blob?.data) {\n const mime = blob.mimeType || blob.mime_type;\n const suffix = ({"image/png":"png", "image/jpeg":"jpg", "image/webp":"webp", "audio/wav":"wav", "audio/mpeg":"mp3", "video/mp4":"mp4"})[mime] || "bin";\n saveBlob(blob.data, suffix, `candidate-${ci}-part-${pi}`);\n }\n }\n }\n}\n\nsaveResult(resp);\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='deepseek-ai/deepseek-v4.1-flash',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 yoki SSE
Qaytarilgan har bir to‘liq javob, fayl va kontent qismini saqlang.
Har bir javob variantini va uning kontenti, vosita chaqiruvlari hamda fikrlash ma’lumotlarini saqlang. SSEda o‘zgarish qismlarini variant va vosita indeksi bo‘yicha yig‘ing, yakuniy tugash sababi va sarf ma’lumotlarini saqlang.
So‘rov va javob IDlari hamda sarlavhalarini saqlang. Matn vazifasini olish uchun umumiy URL taqdim etilmagan.
| Maydon | Tur |
|---|---|
Maydonchoices[].index | Turinteger |
Maydonchoices[].message.content | Turstring|null |
Maydonchoices[].message.tool_calls | Turtool_call[] |
Maydonchoices[].message.reasoning_content | Turstring |
Maydonchoices[].message.reasoning_details | Turobject|array |
Maydonchoices[].finish_reason | Turstring|null |
Maydonusage | Turobject |
| Javob | Shart | Keyingi qadam |
|---|---|---|
Javob400 · invalid_request_error | ShartMajburiy maydonlar, material ifodasi yoki parametr qiymati noto‘g‘ri | Keyingi qadamYangi yaratish so‘rovidan oldin ko‘rsatilgan maydonlarni tuzating. |
Ailoot deepseek-ai/deepseek-v4.1-flash
Yuqorida ko‘rsatilgan usul yo‘llaridan foydalaning.
Ingliz tilidagi texnik misol
Trace a data-loss bug from a text log\n\nUse this log and code sketch to identify the likely lost-update mechanism: A reads stock=5; B reads stock=5; A writes 4; B writes 4. Both requests call read(), subtract one, then write(). Explain the failure, propose one atomic update strategy and a concurrent test. Do not assume a specific database or execute anything.A causal diagnosis, database-neutral pseudocode for an atomic decrement and a concurrent-test scenario with the expected final count.The test expects stock 3 after two successful decrements from 5, with a stated insufficient-stock rule.