Manyan Masu Samar da API na LLM 12 a 2026 (Jagorar ShareAI)

An sabunta a watan Fabrairu 2026 · ~12 minti karantawa

Masu samar da API na LLM 2026 sun fi muhimmanci fiye da kowane lokaci don aikace-aikacen samarwa. Kuna buƙatar ingantaccen, mai tsada mai rahusa wanda ke daidaitawa, lura da abin da ke kiyaye ku gaskiya, da 'yancin jagorantar zirga-zirga zuwa mafi kyawun samfurin don kowanne aiki—ba tare da kullewa ba.

Wannan jagorar yana kwatanta manyan masu samar da API na LLM guda 12 na 2026 kuma yana nuna inda RabaAI ya dace ga ƙungiyoyin da ke son API ɗaya mai dacewa da OpenAI, hanyar jagoranci da mutane ke sarrafawa a cikin samfura 150+, da ganuwar farashi & jinkiri da aka gina—don haka za ku iya jigilar kaya cikin sauri da kashewa cikin hikima. Don gano samfurin, duba Kasuwar Samfura kuma fara gina tare da Bayanin API.

Me yasa Masu Samar da API na LLM 2026 Suke da Mahimmanci

Daga samfurin farko zuwa samarwa: inganci, jinkiri, farashi, sirri

Inganci: zirga-zirgar samarwa yana nufin tsalle-tsalle, sake gwadawa, madadin, da tattaunawar SLA—ba kawai hanyar gwaji mai kyau ba.

Jinkiri: lokaci-zuwa-tokens-na-farko (TTFT) da tokens/sec suna da mahimmanci ga UX (tattaunawa, wakilai) da farashin kayan aiki (lokacin lissafi da aka adana).

Farashi: tokens suna taruwa. Jagorantar zuwa samfurin da ya dace ga kowanne aiki na iya rage kashewa da kashi biyu cikin goma a sikeli.

Sirri & bin cike da doka: sarrafa bayanai, zama a yankin, da manufofin riƙewa sune muhimman abubuwa don siye.

Abin da siye ke damuwa da shi vs. abin da masu gini ke buƙata

Siye: SLAs, rajistan dubawa, DPAs, SOC2/HIPAA/ISO tabbaci, yankuna, da tsinkayar farashi.

Masu gini: faɗin samfurin, TTFT/tokens-per-second, kwanciyar hankali na yawo, tagogin mahallin, ingancin embeddings, gyaran kai, da sauyawa samfurin ba tare da matsala ba. Bincika Gida na Takardu kuma Filin wasa.

Matsayi TL;DR—kasuwa vs. mai bayarwa guda ɗaya vs. ShareAI

APIs na mai bayarwa guda ɗaya: kwangiloli masu sauƙi; zaɓin samfurin da aka iyakance; yiwuwar farashi mai tsada.

Kasuwanni/masu tura: samfurori da yawa ta hanyar API guda ɗaya; siyayya farashi/ayyuka; sauyawa tsakanin masu bayarwa.

ShareAI: kasuwa mai ƙarfin mutane + lura ta tsohuwa + dacewa da OpenAI + babu kullewa.

Masu bayar da API na LLM 2026: Kwatanta a Taƙaice

Waɗannan hotunan jagora ne don taimakawa rage zaɓuɓɓuka. Farashi da nau'ikan samfura suna canzawa akai-akai; tabbatar da kowanne mai bayarwa kafin yanke shawara.

Mai bayarwa	Nau'in Farashi Na Al'ada	Halayen Jinkiri (TTFT / Throughput)	Taga Mahallin (na al'ada)	Faɗi / Bayanan kula
ShareAI (na'ura mai ba da hanya)	Ya bambanta da mai bayarwa da aka tsara; bisa manufofi (farashi/jinkiri)	Ya dogara da hanyar da aka zaɓa; auto-failover & zaɓin yanki	Ya dogara da mai bayarwa	Sama da samfura 150; Mai jituwa da OpenAI; lura da ginannen tsarin; hanyar manufofi; failover; BYOI ana tallafawa
Tare AI	Ta token ta samfurin	Ƙararrakin ƙasa da 100ms akan ingantattun tsaruka	Har zuwa 128k+	200+ samfuran OSS; gyaran ƙarshe
Wutar wutsiya AI	Per-token; ba tare da sabar ba & akan buƙata	Ƙaramin TTFT sosai; ƙarfi mai yawa	128k–164k	Rubutu+hoto+sauti; FireAttention
OpenRouter (na'ura mai ba da hanya)	Na musamman ga samfurin (ya bambanta)	Ya dogara da mai bayarwa na asali	Na musamman ga mai bayarwa	~300+ samfura ta hanyar API ɗaya
Hyperbolic	Ƙananan per-token; mayar da hankali akan rangwame	Saurin shigar da samfurin	~131k	API + GPUs masu araha
Maimaitawa	Amfani da kowane bincike	Ya bambanta da samfurin al'umma	Na musamman ga samfurin	Samfura masu tsawo; protos masu sauri
Hugging Face	APIs masu masauki / masaukin kai	Mai dogaro da kayan aiki	Har zuwa 128k+	OSS hub + gadoji na kamfanoni
Groq	Na kowane token	TTFT mai matukar ƙasa (LPU)	~128k	Gudanar da hanzari ta hanyar kayan aiki
DeepInfra	Kowane-token / na musamman	Tsayayyen fassara a matakin girma	64k–128k	Ana samun wuraren ƙarshe na musamman
Rikicewa (pplx-api)	Amfani / biyan kuɗi	An inganta don bincike/QA	Har zuwa 128k	Saurin samun sabbin samfuran OSS
Anyscale	Amfani; kasuwanci	Sikelin Ray-naƙasasshe	Dogaro mai dogaro da nauyi	Dandalin ƙarshen-zuwa-ƙarshen akan Ray
Novita AI	Kowane-token / kowane-sakanni	Ƙananan farashi + farawa mai sauri	~64k	Ba tare da uwar garke + GPUs na musamman

Lura kan hanyar: rahoton TTFT/tokens/sec yana bambanta da tsawon tambaya, caching, batching, da wurin uwar garke. Yi la'akari da lambobi a matsayin alamomin dangantaka, ba cikakkun bayanai ba. Don samun hanzari Masu samar da API na LLM 2026, kwatanta farashi, TTFT, tagogin mahallin, da faɗin samfurin a sama.

Inda ShareAI Ya Dace Tsakanin Masu Bayar da API na LLM 2026

Kasuwar da mutane ke sarrafawa: samfura 150+, hanyoyin sassauƙa, babu kullewa

ShareAI yana haɗa manyan samfura (OSS da mallakar) a bayan API mai dacewa da OpenAI. Yi hanyar kowane buƙata ta sunan samfurin ko ta manufofi (mafi arha, mafi sauri, mafi daidaito don aiki), ta atomatik canza lokacin da yankin ko samfurin ya samu matsala, kuma canza samfura da layi ɗaya—ba tare da sake rubuta aikinku ba. Ziyarci Bayanin Console.

Kulawar farashi & lura ta tsohuwa

Samu lokaci-lokaci na ainihi, jinkiri, kuskure, da bin diddigin farashi a matakin buƙata da mai amfani. Raba ta mai bayarwa/model don gano koma baya da inganta manufofin hanya. Rahoton da ya dace da siye ya haɗa da yanayin amfani, tattalin arziki na raka'a, da hanyoyin bincike. Tsakanin Masu samar da API na LLM 2026, ShareAI yana aiki a matsayin matakin sarrafawa tare da hanyoyin hanya, failover, lura, da BYOI.

API ɗaya, masu bayarwa da yawa: rashin gogayya na sauyawa

ShareAI yana amfani da keɓaɓɓen OpenAI don haka za ku iya ci gaba da amfani da SDKs ɗinku. Takardun izini suna ci gaba da kasancewa a iyakance; kawo mabuɗan ku inda ake buƙata. Babu kullewa: tambayoyinku, rajistar ku, da manufofin hanyoyin ku suna ɗaukar nauyi. Lokacin da kuka shirya jigilar kaya, duba sabbin Lura Kan Saki.

Gwada shi cikin mintuna 5 (lambar farko ta mai gini)

curl -s https://api.shareai.now/api/v1/chat/completions \"

Don gwaji Masu samar da API na LLM 2026 ba tare da sake fasalin ba, hanya ta hanyar ShareAI’s OpenAI-compatible endpoint da ke sama kuma kwatanta sakamako a ainihin lokaci.

Yadda Za a Zaɓi Mai Bayar da API na LLM da Ya Dace (2026)

Matrix na yanke shawara (jinkiri, farashi, sirri, sikeli, samun samfurin)

Tattaunawa/agents masu mahimmanci ga jinkiri: Groq, Fireworks, Together; ko ShareAI yana hanyar zuwa mafi sauri a kowace yanki.

Tsada mai mahimmanci na rukuni: Hyperbolic, Novita, DeepInfra; ko ShareAI manufofin da aka tsara don rage tsada.

Bambancin samfur / sauyawa cikin sauri: OpenRouter; ko ShareAI mai samarwa da yawa tare da failover.

Gudanarwa na kamfanoni: Anyscale (Ray), DeepInfra (na musamman), tare da rahotanni & bincike na ShareAI.

Multimodal (rubutu+hoto+audio): Fireworks, Together, Replicate; ShareAI na iya jagorantar su. Don tsari mai zurfi, fara a Gida na Takardu.

Takaitaccen jerin kungiyoyi Masu samar da API na LLM 2026 ya kamata su gwada a yankin da suke yi wa hidima don tabbatar da TTFT da tsada.

Ayyuka: manhajojin hira, RAG, wakilai, rukuni, multimodal

UX na hira: fifita TTFT da tokens/sec; kwanciyar hankali na yawo yana da mahimmanci.

MAZA: ingancin embeddings + girman taga + tsada.

Wakilai/kayan aiki: ƙarfafa kira-aiki; sarrafa lokaci; sake gwadawa.

Batch/kan layi: yawan aiki da $ a kowane 1M alamu sun mamaye.

Multimodal: samuwar samfurin da farashin alamu marasa rubutu.

Jerin bincike na siye (SLA, DPA, yanki, riƙe bayanai)

Tabbatar da burin SLA da kuɗaɗen, sharuɗɗan DPA (sarrafawa, ƙananan masu sarrafawa), zaɓin yanki, da manufofin riƙe don tambayoyi/fitarwa. Nemi hanyoyin lura (headers, webhooks, fitarwa), sarrafa bayanai na gyara, da zaɓuɓɓukan BYOK/BYOI idan an buƙata. Duba Jagorar Mai Samarwa idan kuna shirin kawo ƙarfin aiki.

Manyan masu bayar da API LLM 12 na 2026

Kowanne bayanin martaba ya haɗa da taƙaitaccen “mafi dacewa”, dalilin da yasa masu gini suka zaɓe shi, farashi a taƙaice, da bayanai kan yadda yake dacewa tare da ShareAI. Waɗannan su ne Masu samar da API na LLM 2026 waɗanda aka fi kimantawa don samarwa.

1) ShareAI — mafi dacewa don hanyar sadarwa mai yawa, lura & BYOI

Dalilin da yasa masu gini suka zaɓe shi: API ɗaya mai dacewa da OpenAI a cikin samfura 150+, hanyar sadarwa mai dogaro da manufofi (farashi/lokaci/daidaito), auto-failover, nazarin farashi & lokaci na ainihi, da BYOI lokacin da kuke buƙatar ƙarfin aiki na musamman ko sarrafa bin doka.

Farashi a taƙaice: yana bin farashin mai ba da hanya; kuna zaɓar manufofin da aka inganta farashi ko waɗanda aka inganta jinkiri (ko wani mai ba da sabis/misali na musamman).

Bayanan kula: madaidaicin “matakin sarrafawa” ga ƙungiyoyi da ke son 'yanci don canza masu ba da sabis ba tare da sake fasalin tsarin ba, kiyaye siyan kayayyaki cikin farin ciki tare da rahotannin amfani/kuɗi, da gwaji a cikin samarwa.

2) Together AI — mafi kyau don manyan LLMs na buɗaɗɗen tushe

Dalilin da yasa masu gini suka zaɓe shi: farashi/ayyuka masu kyau akan OSS (misali, ajin Llama-3), tallafin gyare-gyare, da'awar ƙasa da 100ms, babban kundin adireshi.

Farashi a taƙaice: ta token ta samfurin; ana iya samun kuɗin kyauta don gwaje-gwaje.

Dacewar ShareAI: hanya ta tare/<model-id> ko barin manufofin ShareAI da aka inganta farashi su zaɓi Together idan ya fi arha a yankinku.

3) Fireworks AI — mafi kyau don ƙarancin jinkiri mai yawa

Dalilin da yasa masu gini suka zaɓe shi: TTFT mai sauri sosai, injin FireAttention, rubutu+hoto+sauti, zaɓuɓɓukan SOC2/HIPAA.

Farashi a taƙaice: biya-yadda-kuke-amfani (ba tare da sabar ba ko akan buƙata).

Dacewar ShareAI: kira wuta-wutar/<model-id> kai tsaye ko barin tsarin manufofi ya zaɓi Fireworks don tambayoyin multimodal.

4) OpenRouter — mafi kyau don samun dama ga masu samarwa da yawa ta hanyar API ɗaya

Dalilin da yasa masu gini suka zaɓe shi: ~300+ samfura a bayan API ɗaya; mai kyau don binciken samfurori cikin sauri.

Farashi a taƙaice: farashin kowane samfur; wasu matakai na kyauta.

Dacewar ShareAI: ShareAI yana rufe wannan bukatar mai samarwa da yawa amma yana ƙara tsarin manufofi + lura + rahotanni masu daraja na siye.

5) Hyperbolic — mafi kyau don adana kuɗi sosai & saurin fitar da samfurori

Dalilin da yasa masu gini suka zaɓe shi: farashin kowane token mai sauƙi, saurin farawa don sabbin samfuran buɗaɗɗen tushe, da samun damar GPUs masu araha don ayyuka masu nauyi.

Farashi a taƙaice: kyauta don farawa; biyan kuɗi kamar yadda ake amfani.

Dacewar ShareAI: nuna zirga-zirga zuwa hyperbolic/ don gudanarwa mafi ƙarancin farashi, ko saita manufofi na musamman (misali, “farashi-sannan-lokaci”) don ShareAI ya fi son Hyperbolic amma ya canza kai tsaye zuwa hanya mafi arha mai lafiya yayin tashi.

6) Replicate — mafi kyau don gwaji & samfuran dogon wutsiya

Dalilin da yasa masu gini suka zaɓe shi: babban kundin al'umma (rubutu, hoto, sauti, samfuran niche), deploy ɗaya don MVPs cikin sauri.

Farashi a taƙaice: kowane bincike; yana bambanta da kwantena samfur.

Dacewar ShareAI: mai kyau don gano; lokacin da ake girma, yi hanya ta ShareAI don kwatanta latency/kudin da madadin ba tare da canje-canje na lambar ba.

7) Hugging Face — mafi kyau don tsarin OSS & gadoji na kamfanoni

Dalilin da yasa masu gini suka zaɓe shi: hub na samfur + datasets; fassarar da aka shirya ko kai tsaye a cikin gajimare naka; gadoji masu ƙarfi na MLOps na kamfanoni.

Farashi a taƙaice: kyauta don abubuwan asali; shirye-shiryen kamfanoni suna samuwa.

Dacewar ShareAI: adana samfuran OSS naka kuma yi hanya ta ShareAI don haɗa HF endpoints tare da sauran masu samarwa a cikin app ɗaya.

8) Groq — mafi kyau don latency mai ƙanƙanta sosai (LPU)

Dalilin da yasa masu gini suka zaɓe shi: fassarar da aka hanzarta ta kayan aiki tare da TTFT/tokens-per-second na masana'antu don hira/agents.

Farashi a taƙaice: per-token; mai dacewa da kamfanoni.

Dacewar ShareAI: amfani groq/<model-id> a cikin hanyoyin da ke da damuwa da latency; saita ShareAI failover zuwa hanyoyin GPU don juriya.

9) DeepInfra — mafi kyau don masaukin da aka keɓe & fassarar mai tsada

Dalilin da yasa masu gini suka zaɓe shi: API mai tsayayye tare da tsarin OpenAI-style; endpoints da aka keɓe don LLMs masu zaman kansu/jama'a.

Farashi a taƙaice: per-token ko lokacin aiwatarwa; farashin na'ura da aka keɓe yana samuwa.

Dacewar ShareAI: mai amfani lokacin da kake buƙatar ƙarfin da aka keɓe yayin kiyaye nazarin masu samarwa ta ShareAI.

10) Perplexity (pplx-api) — mafi kyau don haɗin bincike/QA

Dalilin da yasa masu gini suka zaɓe shi: samun sauri zuwa sabbin samfuran OSS, sauƙaƙan REST API, mai ƙarfi don dawo da ilimi da QA.

Farashi a taƙaice: bisa amfani; Pro yawanci yana haɗawa da kuɗin API na wata-wata.

Dacewar ShareAI: haɗa pplx-api don dawo da bayanai tare da wani mai bayarwa don samarwa a ƙarƙashin aikin ShareAI guda ɗaya.

11) Anyscale — mafi kyau don ƙarshen-zuwa-karshen sikeli akan Ray

Dalilin da yasa masu gini suka zaɓe shi: horo → hidima → tsari akan Ray; fasalolin gudanarwa/tsari don ƙungiyoyin dandamali na kasuwanci.

Farashi a taƙaice: bisa amfani; zaɓuɓɓukan kasuwanci.

Dacewar ShareAI: daidaita kayan aiki akan Ray, sannan amfani da ShareAI a gefen aikace-aikace don hanyar sadarwa ta masu bayarwa da haɗin bayanai.

12) Novita AI — mafi kyau don serverless + GPU na musamman a farashi mai rahusa

Dalilin da yasa masu gini suka zaɓe shi: biyan kuɗi na kowane dakika, farawa mai sauri, hanyar sadarwa ta duniya ta GPU; duka serverless da na musamman.

Farashi a taƙaice: bisa token (LLM) ko bisa dakika (GPU); hanyoyin ƙarshe na musamman don kasuwanci.

Dacewar ShareAI: mai ƙarfi don adana farashin tsari; ci gaba da hanyar ShareAI don juyawa tsakanin Novita da abokan aiki ta yankin/farashi.

Fara Da Sauri: Hanyar Duk Wani Mai Bayarwa Ta ShareAI (Hada da Kulawa)

Misali mai dacewa da OpenAI (cikakkun tattaunawa)

curl -s https://api.shareai.now/api/v1/chat/completions \"

Sauya masu samarwa da layi ɗaya

{
  "model": "growably/deepseek-r1:70b",
  "messages": [
    {"role": "user", "content": "Latency matters for agents—explain why."}
  ]
}

Don gwaji Masu samar da API na LLM 2026 da sauri, kiyaye nauyin kaya iri ɗaya kuma kawai canza samfurin ko zaɓi manufar na'ura mai ba da hanya.

Lura da Gwaji & Ƙayyadaddun abubuwa

Bambance-bambancen Tokenization canza jimlar adadin token tsakanin masu samarwa.

Haɗawa da adanawa na iya sa TTFT ya bayyana ƙasa da gaskiya akan maimaita tambayoyi.

Wurin uwar garke yana da mahimmanci: auna daga yankin da kake yi wa masu amfani hidima.

Talla na taga mahallin ba cikakken labari bane—duba halayen yanke da ingantaccen gudu kusa da iyaka.

Hotunan farashi: koyaushe tabbatar da farashin yanzu kafin yanke shawara. Lokacin da ka shirya, tuntubi Saki kuma Tarihin Blog don sabuntawa.

FAQ: Masu Bayar da LLM API 2026

Menene mai bayar da LLM API?

Wani Mai bayar da LLM API yana ba da damar amfani da manyan samfuran harshe ta hanyar HTTP APIs ko SDKs. Kuna samun damar daidaitawa, sa ido, da SLAs ba tare da kula da GPU ɗinku ba.

Buɗe-tushen vs mallakar: wanne ya fi dacewa don samarwa?

Buɗe-tushen (misali, ajin Llama-3) yana ba da ikon sarrafa farashi, keɓancewa, da sauƙin ɗaukarwa; mallakar samfura na iya jagoranci a wasu ma'auni da sauƙi. Kungiyoyi da yawa suna haɗa duka—RabaAI yana sanya wannan haɗin da daidaitawa mai sauƙi.

Together AI vs Fireworks — wanne ya fi sauri don multimodal?

Wutar wuta an san shi da ƙarancin TTFT da ƙarfi a cikin tsarin multimodal; Tare yana bayar da babban kundin OSS da kuma karfin aiki mai gasa. Zabin ku mafi kyau ya dogara da girman saƙo, yanki, da yanayin aiki. Tare da RabaAI, za ku iya turawa zuwa kowanne kuma ku auna sakamako na gaske.

OpenRouter vs ShareAI — kasuwa vs hanyar turawa ta mutane?

BudeRouter yana haɗa yawancin samfura ta API ɗaya—mai kyau don bincike. RabaAI yana ƙara hanyar turawa bisa manufofi, lura mai dacewa da siye, da kuma tsara ta hanyar mutane don ƙungiyoyi su iya inganta farashi/jinkiri da daidaita rahoto a duk masu samarwa.

Groq vs GPU Cloud — yaushe LPU ke cin nasara?

Idan aikin ku yana da mahimmanci ga jinkiri (wakilai, tattaunawa mai hulɗa, UX mai gudana), Groq LPUs na iya bayar da TTFT/tokens-per-second mafi kyau a masana'antu. Don ayyukan tsari masu nauyi, masu samar da GPU masu inganta farashi na iya zama mafi tattalin arziki. RabaAI yana ba ku damar amfani da duka biyun.

DeepInfra vs Anyscale — keɓaɓɓen fassarar hankali vs Ray dandamali?

DeepInfra yana haskakawa don keɓaɓɓen fassarar hankali; Anyscale dandamali ne na Ray wanda ya haɗa horo zuwa hidima zuwa tsari. Ƙungiyoyi sau da yawa suna amfani da Anyscale don tsara dandamali da RabaAI a gefen aikace-aikace don turawa tsakanin masu samarwa da nazari.

Novita vs Hyperbolic — mafi ƙarancin farashi a sikeli?

Dukansu suna gabatar da tanadi mai tsauri. Novita yana mai da hankali kan serverless + GPUs na musamman tare da biyan kuɗi na daƙiƙa-daƙiƙa; Hyperbolic yana nuna samun GPUs da aka rage farashi da saurin onboarding na samfurin. Gwada duka tare da tambayoyinku; yi amfani da ShareAI’s na'ura mai ba da hanya:farashi_mai_kyau don kiyaye farashi gaskiya.

Replicate vs Hugging Face — prototyping vs zurfin yanayin?

Maimaitawa ya dace don prototyping mai sauri da samfuran al'umma masu tsawon lokaci; Hugging Face yana jagorantar yanayin OSS tare da gadoji na kamfanoni da zaɓuɓɓukan yin masauki da kansa. Yi amfani da hanya ta RabaAI don kwatanta farashi & latency cikin gaskiya.

Wane ne mafi tsadar LLM API mai ba da sabis a 2026?

Ya dogara da haɗin tambaya da yanayin zirga-zirga. Masu fafatawa masu mai da hankali kan farashi: Hyperbolic, Novita, DeepInfra. Hanyar da ta fi dacewa don amsawa ita ce auna tare da RabaAI lura da kuma manufofin hanyar da aka inganta don farashi.

Wane mai bayarwa ne mafi sauri (TTFT)?

Groq sau da yawa yana jagoranci akan TTFT/tokens-per-second, musamman don UX na hira. Wutar wuta kuma Tare kuma suna da ƙarfi. Koyaushe gwada a yankinku—kuma bari RabaAI hanya zuwa mafi saurin ƙarshen kowane buƙata.

Mafi kyawun mai bayarwa don RAG/agents/batch?

MAZA: babban mahallin + ingancin embeddings; yi la'akari Tare/Fireworks; haɗa tare da pplx-api don dawo da bayanai. Wakilai: ƙananan TTFT + kiran aiki mai dogaro; Groq/Fireworks/Tare. Batch: farashi ya yi nasara; Novita/Hyperbolic/DeepInfra. Hanya tare da RabaAI don daidaita sauri da kashe kudi.

Tunani Na Ƙarshe

Idan kana zaɓa tsakanin Masu samar da API na LLM 2026, kada ka zaɓi bisa farashi da labarai kaɗai. Yi gwajin mako 1 tare da tambayoyinka na gaske da bayanan zirga-zirga. Yi amfani da RabaAI don auna TTFT, throughput, kurakurai, da farashin kowane buƙata tsakanin masu samarwa—sannan ka tabbatar da manufar hanya da ta dace da burinka (mafi ƙarancin farashi, mafi ƙarancin jinkiri, ko haɗin hankali). Lokacin da abubuwa suka canza (kuma za su canza), za ka riga ka sami ikon lura da sassauci don sauyawa—ba tare da sake fasalin ba.