Isithombe nguMbhali
# Isingeniso
Ukubhala nge-Vibe kumayelana nokwakha ngokushesha, ukuhlala ugxilile, nokugcina umfutho ngaphandle kokuhlale ucabanga ngemikhawulo yokusetshenziswa noma izindleko.
Uma usebenzisa i-Claude Code nge-API, inkokhelo ingakhula ngokushesha okukhulu. Ukuphindaphinda kaningi, ukulungisa iphutha, nokuhlola kwenza ukuhamba komsebenzi okusekelwe ku-API kubize kumaseshini amade wokubhala amakhodi. Lesi esinye sezizathu eziyinhloko ezenza ukuthi ukubhaliswa kuka-Claude Code Pro kanye no-Max kudume phakathi kwamakhodi we-vibe nonjiniyela, njengoba benikeza ukufinyelela okuqondile kumamodeli ngaphandle kwentengo yesicelo ngasinye.
Lezi zinhlelo ziza nemikhawulo yokusetshenziswa esethwa kabusha ngemva kwamahora amane, futhi kwezinye izimo ihlanganisa nemikhawulo yamasonto onke. Lokhu kuzenza zibikezeleke kakhulu futhi zilungele izikhathi ezinde, ezingaphazamiseki zokubhala amakhodi.
Kulesi sihloko, sizohlola amapulani ayisikhombisa aphezulu okubhala amakhodi atholakalayo namuhla, ukuthi ipulani ngalinye linikeza ini, nokuthi yiluphi uhlobo lukamakhi noma unjiniyela olulunge kakhulu.
# 1. Claude Code Plans
Izinhlelo ze-Claude Code yilapho ukubhaliswa kwekhodi ye-AI okubikezelwayo kwaqala khona ngempela. Njengoba abathuthukisi beqala ukusebenzisa u-Claude ngezikhathi zokubhala amakhodi ezinde neziphindaphindwayo, i-Claude API yasheshe yabiza kakhulu ukuthi isetshenziswe ngokuqhubekayo.
Ukukhokha ithokheni ngayinye kwenze kwaba nzima ukuzama ngokukhululekile, ukufaka ikhodi kabusha, noma ukuhlala endleleni yokudala. Ukuze kuxazululwe lokhu, i-Anthropic yethula izinhlelo zokubhaliselwe ezihlanganisa ukufinyelela kwe-Claude Code kuma-tier anyanga zonke ahambisana nokusetha kabusha ukusetshenziswa kwamahora amahlanu kanye nemikhawulo eyengeziwe yangeviki ezinhlelweni eziphakeme.
Le ndlela yenza izikhathi ezinwetshiwe zokufaka amakhodi ukuthi zithengeke futhi zilawuleke, futhi yasungula imodeli elandelwa izinhlelo eziningi zesimanje zokubhala amakhodi ze-AI manje.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| UClaude Pro | 20 | Cishe 10 kuya ku-40 ukwaziswa kwekhodi ye-Claude ngamahora angu-5 |
| U-Claude Max (5×) | 100 | Cishe iziyalezo ezingama-50 kuye kwezingama-200 ngamahora ama-5 |
| U-Claude Max (20×) | 200 | Cishe ukwaziswa okungu-200 kuya ku-800 ngamahora angu-5 |
Ukusetshenziswa kusetha kabusha njalo ngamahora amahlanu. Uphahla lwamasonto onke lungasebenza ngisho noma iwindi lamahora amahlanu lingasetshenziswa ngokugcwele.
# 2. I-ChatGPT Codex Plans
Izinhlelo ze-ChatGPT Codex yindlela i-OpenAI ehlanganisa ngayo Amakhono ekhodi ye-Codex ngaphakathi kokubhaliselwe okujwayelekile kwe-ChatGPT, ehlinzeka ngemikhawulo yokusetshenziswa ehlelekile kunezintengo ezikhokha njengoba uhamba.
I-Codex ifakwe nezinhlelo ze-ChatGPT Plus, Pro, Business, kanye ne-Enterprise, futhi lezi zigaba zilawula ukuthi mingaki imilayezo ongayithumela esikhathini esithile kanye nokuthi kungakanani ukubhala ngekhodi ongakwenza ngaphambi kokuba imikhawulo iqale ukusebenza.
Imikhawulo yokusetshenziswa iyahlukahluka ngohlelo futhi ingase isethwe kabusha ngewindi lesikhathi esinqunyiwe, okwenza kube lula kubathuthukisi ukuhlela izikhathi ezinde zokubhala ikhodi uma kuqhathaniswa nenkokhiso esekelwe ku-API.
Lezi zinhlelo ezihlelekile zisize ukusungula indlela ebikezelwe nethengekayo yokwakha nge-Codex ngaphakathi kwe-ChatGPT kubasebenzisi abaningi.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| I-ChatGPT Plus | 20 | Cishe imiyalezo engama-30 kuye kwengama-150 ngamahora ama-5 |
| I-ChatGPT Pro | 200 | Cishe imilayezo engu-300 kuya ku-1500 ngamahora angu-5 |
| Ibhizinisi le-ChatGPT | ~30 umsebenzisi ngamunye | Amakepisi aphezulu womsebenzisi ngamunye, amawindi amahora amahlanu |
| I-ChatGPT Enterprise | Ngokwezifiso | Izilinganiso zangokwezifiso |
Imikhawulo yemilayezo iyahlukahluka ngemodeli nobunkimbinkimbi bemiyalezo.
# 3. I-Google AI Plans
I-Google AI ihlela ukukhuphula imikhawulo yokusebenzisa Usizo lwekhodi ye-Gemini futhi I-Gemini CLI ngokunikeza ababhalisile ama-quota aphezulu wansuku zonke nokufinyelela okubalulekile kumamodeli namathuluzi anamandla.
Ngokungafani nezinye izinhlelo zokufaka amakhodi ezisetha kabusha imikhawulo kumafasitela omfushane, i-Google AI Pro kanye ne-Ultra isebenzisa imikhawulo ikakhulukazi nsuku zonkeokusho ukuthi ungasebenzisa isabelo sakho usuku lonke ngaphandle kokukhathazeka ngokusetha kabusha okufushane.
Ababhalisile abanalezi zinhlelo bathola ngokuzenzakalelayo imikhawulo yesicelo ekhushuliwe yansuku zonke yokugeleza komsebenzi wokubhala ngokuqhathaniswa nama-akhawunti wamahhala, okwenza izikhathi ezinde nemisebenzi yokuthuthukisa esindayo isebenziseke kakhulu futhi ibikezelwe kunokuthembela emikhawulweni yezigaba zamahhala kuphela.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| I-Google AI Pro | ~20 | Cishe izicelo zokufaka amakhodi ezingama-500 kuye kwezingu-1,500 ngosuku kuyo yonke i-Gemini Code Assist ne-Gemini CLI |
| I-Google AI Ultra | ~250 | Izicelo zekhodi ezingaba ngu-3,000 kuya kweziyi-10,000 ngosuku ezinokufinyelela okubaluleke kakhulu |
Imikhawulo yokusetshenziswa isetshenziswa ngokuyinhloko nsuku zonke. Izilinganiso eziqondile zingahluka ngethuluzi, inguqulo yemodeli, nobunkimbinkimbi besicelo, futhi i-Google ingase ilungise imikhawulo ngaphandle kwesaziso somphakathi.
# 4. Izinhlelo Zokufaka Ikhodi ze-GLM
Izinhlelo Zokufaka Amakhodi ze-GLM zinikeza enye yezindlela ezingabizi kakhulu nezivumelana nezimo zokwenza ukubhala ngekhodi okusizwa yi-AI ngokuhlanganisa izibalo zokwaziswa zibe yizigaba ezimisiwe zanyanga zonke ezisetha kabusha njalo ngemva kwamahora amahlanu.
Lezi zinhlelo zakhelwe ukugeleza komsebenzi wokufaka amakhodi oshayelwa yi-ejenti futhi zinikeza onjiniyela izilinganiso ezibikezelwe kuwo wonke amathuluzi adumile njenge-Claude Code, i-Cline, ne-OpenCode ngaphandle kwezindleko eziphezulu zethokheni lokunye okubhaliselwe.
Esigabeni esiphansi kunazo zonke, uhlelo luqala endaweni ezungezile 3 amadola ngenyanga futhi isivele inikezela ngomthamo owanele wokusekela izikhathi zokubhala amakhodi njalo, kuyilapho izigaba eziphakeme zikala kakhulu kunalokho ukuze zinikeze izidingo zentuthuko ezidingeka kakhulu.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| I-GLM Lite | ~3 | Cishe ukwaziswa kwe-120 ngamahora angu-5 |
| I-GLM Pro | ~15 | Cishe ukwaziswa okungu-600 ngamahora angu-5 |
| Inani eliphakeme kakhulu lama-GLM | ~30 | Cishe ukwaziswa okungu-2,400 ngamahora amahlanu |
Izibalo zokwaziswa zisethwa kabusha njalo ngemva kwamahora amahlanu, okunikeza onjiniyela iwindi elibikezelwe lokubhala, ukulungisa iphutha, nokuphindaphinda kukhodi.
# 5. MiniMax Coding Plans
I-MiniMax Coding Plans inikezela ngesakhiwo samanani esicacile nesicacile kakhulu sokufakwa kwekhodi kwe-AI, okuzenza zithandeke kakhulu kubathuthukisi abafuna ama-quota angabikezelwa ngaphandle kwezindleko eziphezulu ze-API.
Isigaba ngasinye sinikeza inombolo egxilile yokwaziswa ngaphakathi kwewindi eligoqayo lamahora amahlanu, futhi ukwaziswa okukodwa kuhamba kakhulu kunokwaziswa okukodwa kumodeli eyisisekelo ngoba kungamela izicelo eziningi ngaphakathi.
Lezi zinhlelo zinikwa amandla imodeli ye-MiniMax M2.1, eklanyelwe ukubhala amakhodi okuphumelelayo nokugeleza komsebenzi we-ejenti, futhi inika onjiniyela ukulawula okukhulu izindleko nokusetshenziswa kunezinye izindlela zokukhokha njengoba uhamba.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| I-MiniMax Starter | 10 | 100 ukwaziswa ngamahora angu-5 |
| I-MiniMax Plus | 20 | 300 ukwaziswa ngamahora angu-5 |
| I-MiniMax Max | 50 | 1000 ukwaziswa ngamahora angu-5 |
Izibalo zokwaziswa zisethwa kabusha njalo ngemva kwamahora amahlanu, okunikeza onjiniyela amawindi acacile, abikezelwe ukuthi babhale, balungise iphutha, futhi baphindaphinde kukhodi ngaphandle kokukhathazeka ngokukhokhiswa kwe-API okungalindelekile.
# 6. Izinhlelo ze-Kimi Coding
I-Kimi Coding Plans ifakwe nobulungu be-Kimi futhi ihlinzeka ngezilinganiso zesicelo sokubhala ikhodi masonto onke esikhundleni samawindi e-sprint amafushane.
Uma ubhalisa, uthola inombolo emisiwe yezicelo zamakhodi zeviki ezivuselelwa njalo ezinsukwini eziyisikhombisa kusukela osukwini lwakho lokusebenza, futhi isabelo esingasetshenzisiwe asidluli umjikelezo wamasonto onke.
Izilinganiso eziqondile zezinombolo azishicilelwa esidlangalaleni, kodwa imibiko yabasebenzisi nezithenjwa zedeshibhodi ziphakamisa ukuthi amalungu e-Starter angase abone ngokulandelana kwezicelo ezingu-2,000 kuya kwezingu-3,500 ngesonto, kuyilapho amalungu e-Pro noma e-Ultra ethola izibonelelo zeviki ezinkulu kakhulu.
Lolu hlelo lwesabelo lwamasonto onke lwenza izinhlelo zibikezeleke konjiniyela ababhala amakhodi njalo iviki lonke kunokuba kube ngamaphutha amafushane.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| Kimi Ubulungu Starter | 9 ngo10 | ~2,000 kuya ku-3,500 izicelo zokufaka amakhodi ngesonto |
| Kimi Ubulungu Pro noma Ultra | ~ 49 | ~8,000 kuya ku-15,000 izicelo zokufaka amakhodi ngesonto |
Izabelo zivuselelwa kumjikelezo wezinsuku eziyisikhombisa oqala kusukela ekwenzeni kusebenze okubhaliselwe. Imikhawulo yezinombolo eqondile iyabonakala kudeshibhodi yomsebenzisi kodwa ayishicilelwa njengezinombolo zomphakathi ezingashintshi.
# 7. Cerebras Code Plans
Cerebras Code Plans yakhelwe onjiniyela abadinga ukukhiqiza okuphezulu kakhulu nesivinini kokuhamba komsebenzi wokufaka amakhodi kwe-AI. Esikhundleni sokukhawulela inani lezaziso noma imilayezo, i-Cerebras isebenzisa imikhawulo ngokuyinhloko amathokheni ngosukuenikeza ababhalisile izibonelelo zansuku zonke ezinkulu ezisekela ukubhala ngekhodi okuqhubekayo, okuqhubekayo kunamafasitela omfushane.
Ngokufinyelela ku i-hardware ye-inference esheshayo egijima efika kumathokheni angaba ngu-2,000 ngomzuzwana kanye nama-quota amathokheni amakhulu wansuku zonke, lezi zinhlelo ziphakathi kwezinketho eziphezulu kakhulu ezitholakalayo zokubhala ikhodi ye-vibe kanye nemisebenzi yokuthuthukisa eqhutshwa yi-ejenti esindayo.
| Hlela | Intengo yanyanga zonke (USD) | Imikhawulo yokusetshenziswa |
|---|---|---|
| Cerebras Code Pro | 50 | 24 million amathokheni ngosuku |
| Cerebras Code Max | 200 | 120 million amathokheni ngosuku |
| Imodeli | Cishe. Isivinini (amathokheni ngomzuzwana) |
|---|---|
| I-ZAI GLM 4.7 | ~1,000 |
| I-OpenAI GPT-OSS 120B | ~3,000 |
Izinhlelo ze-Cerebras Code zivumela abathuthukisi ukuthi bakhiqize futhi bahlele ikhodi ngokuqhubekayo usuku lonke ngezabelomali zamathokheni amakhulu kanye nokunye kokusebenza okuphezulu okuqhubekayo embonini.
# Ukuqhathanisa Okulula Kwezinhlelo Zokubhala Ngekhodi ze-AI ezidumile
Leli thebula linikeza ukuqhathanisa okusheshayo kwezinhlelo zokufaka amakhodi ze-AI ezidumile ngokusekelwe enanini, imikhawulo encane esebenzisekayo, nokuthi ukusetshenziswa kusetha kabusha kanjani, ukuze ubone kalula ukuthi iyiphi inketho elingana nesitayela sakho sokubhala.
| Umhlinzeki | Intengo yanyanga zonke (USD) | Isibonelelo Esincane Sokusetshenziswa | Setha kabusha Isitayela | Kuhle kakhulu |
|---|---|---|---|---|
| Claude Ikhodi | 20 kuya ku-200 | ~ 10 ukwaziswa ngamahora angu-5 | Ukugoqa amahora angu-5 kanye namakepisi weviki | Amaseshini amade okubhala amakhodi |
| I-ChatGPT Codex | 20 kuya ku-200+ | ~30 imilayezo ngamahora angu-5 | Ukuhamba amahora angu-5 | Ukubhala amakhodi okujwayelekile kanye nokulungisa iphutha |
| I-Google AI | ~20 kuya ku-250 | ~ Izicelo ezingama-500 ngosuku | Ukusetha kabusha kwansuku zonke | Ukubhala amakhodi okuqinile kwansuku zonke |
| I-GLM | ~ 3 kuya ku-30 | ~ 120 ukwaziswa ngamahora angu-5 | Ukuhamba amahora angu-5 | Inani elishibhile nelingcono kakhulu lekhodi ye-vibe |
| I-MiniMax | 10 ku50 | 100 ukwaziswa ngamahora angu-5 | Ukuhamba amahora angu-5 | Ukubhala ikhodi kwe-vibe okusekelwe ku-Sprint |
| Kimi | 10 kuya ku-49 | ~ Izicelo eziyi-2,000 ngeviki | Isabelo samasonto onke | Ukubhala amakhodi kwamasonto onke okungaguquki |
| I-Cerebras | 50 kuya ku-200 | 24 million amathokheni ngosuku | Ukusetha kabusha kwansuku zonke | Isivinini esikhulu nokubhalwa kwekhodi okuqhubekayo |
Abid Ali Awan (@1abidiawan) uchwepheshe wesayensi yedatha othanda amamodeli wokufunda womshini wokwakha. Njengamanje, ugxile ekudaleni okuqukethwe nasekubhaleni amabhulogi ezobuchwepheshe ekufundeni komshini kanye nobuchwepheshe besayensi yedatha. U-Abid uneziqu ze-Master's in technology management kanye neziqu ze-bachelor's in telecommunication engineering. Umbono wakhe uwukwakha umkhiqizo we-AI esebenzisa i-graph neural network yabafundi abanenkinga yokugula ngengqondo.
