Guudmar

Xakamaynta dhererka jawaabta noocu waxay faa’iido u leedahay dhowr sababood: waxay ka caawisaa maareynta kharashka (maadaama aad token kasta lacag ku bixiso), waxay hagaajisaa daahitaanka/waxqabadka (jawaabaha gaagaaban si dhakhso leh ayaa loo soo celiyaa), waxayna hubisaa khusaynta iyadoo laga fogaanayo wax-soo-saar aad u dheer ama hadal badan.

Waxaad tan ku gaari kartaa adigoo adeegsanaya xadadka token-ka, dejimaha caqliyeynta iyo faahfaahsanaanta, tilmaamo cad, tusaalooyin, iyo taxanayaal joojin. Faahfaahinta ugu cusub uguna dhammaystiran, had iyo jeer tixraac tixraaca rasmiga ah ee API-ga ee platform.openai.com.

Deji dhererka ugu badan ee wax-soo-saarka

API-ga Responses

Waxaa loo adeegsadaa noocyada GPT-5 iyo inta badan noocyada o-series: isticmaal max_output_tokens si aad u xaddiddo tirada token-ka ee noocu abuuri doono. Codsiyada compaction_trigger, ama ka tag max_output_tokens ama u deji ugu yaraan 20000; qiimayaal ka yar waa la diidaa. API-ga Responses ma taageero dhammeystirro badan (n).

API-ga dhammeystirka wada-sheekeysiga

Waxaa loo adeegsadaa GPT-3.5-kii hore, GPT-4o, iyo mararka qaar o-series.

• Noocyada caqliyeynta sida o3 iyo o4-mini, isticmaal max_completion_tokens (magac kale oo max_tokens ah)

• Noocyada hore/kuwa aan caqliyeyn ahayn, max_tokens weli wuu shaqeeyaa

• Waxay taageertaa stop iyo n (dhammeystirro badan).

Xaddidaadaha token-ka marka loo eego kooxaha noocyada

Si aad u hesho xaddidaadaha token-ka ee ugu dambeeyay, cabbirrada macnaha guud, iyo xadadka soo-saarka, fadlan tixraac dukumentiyada nooca gaarka ah.

Tusaalooyin degdeg ah

API-ga Responses

Dhammeystirka Wada-sheekeysiga (nooca caqliyeynta)

Xakamaynnada gaarka u ah noocyada GPT-5: verbosity iyo reasoning.effort

Xakamaynnadan waxaa kaliya laga heli karaa noocyada GPT-5 (gpt-5.2, gpt-5.2-chat-latest, gpt-5.2 pro, iwm.). Noocyada O-series iyo noocyadii hore ma taageeraan.

`verbosity` wuxuu aqbalaa "low", "medium" (caadiga), ama "high". Waxay saamayn ku yeelataa heerka faahfaahinta, balse ma dejiso xadad adag.

`reasoning.effort` wuxuu xakameeyaa tirada token-ka caqliyeynta ee la abuuro ka hor inta aan jawaab la bixin. GPT-5.2 wuxuu taageeraa none,low, medium, high,and xhigh. gpt-5.2-pro wuxuu taageeraa oo keliya medium, high,and xhigh. Noocyadii hore ee caqliyeynta waxay taageeraan oo keliya low, medium, iyo high.

Waxaad `reasoning.effort` u dejin kartaa none si noocu ugu dhaqmo sida nooc aan caqliyeyn ahayn, adeegsiyada daahitaanku muhiimka yahay.

Bixi tilmaamo gaar ah

Codso dhererka ama qaabka saxda ah ee aad rabto. Tusaalooyin:

• “Tax si sax ah shan ikhtiyaar.”

• “Qor soo-koobid 50 eray ah.”

• “Aan ka badnaan 100 token. Haddii aad u baahato wax ka badan, dheh ‘Waxaan u baahanahay boos dheeraad ah.’”

Isticmaal tusaalooyin leh dherer joogto ah

Tusaalooyinka tiro yar ee la jaanqaadaya dhererka aad rabto waxay nooca ka caawiyaan inuu sii wado qaabka.

Ku dabaq taxanayaal joojin istaraatiiji ah

Isticmaal stop si aad u joojiso soo-saarista marka noocu gaaro kala-sooce ama xadka liis lambaraysan.

Ikhtiyaarro badan

• Dhammeystirka Wada-sheekeysiga: n waxay hal wicitaan ku soo celisaa dhammeystirro badan.

• API-ga Responses: n lama taageero; samee wicitaanno badan haddii aad u baahan tahay wax ka badan hal wax-soo-saar.