Opus 5.5 Maalin Cad Ayuu Dhaafay GPT-6-Astra
Sharaxaad
Videogan waxaan kaga hadlayaa Opus 5.5, model-ka cusub ee ay shirkadda Anthropic soo saartay. Anthropic waxay sheegtay in model-kani shaqada inteeda badan kula siman yahay Fable, isla markaana uu 40% ka jaban yahay Opus-kii hore. Dhibaatada ugu weyn ee models-ka Opus waxay ahayd token usage-ka: task-ga aad u dirto malaayiin token ayuu ku bixin jiray, lacag badanna wuu kaa dhammayn jiray. Benchmark-yada agentic coding-ka wuxuu helay 66.4 halka Fable uu ku jiro 55, Terminal-Bench-na wuxuu gaaray 54 halkii uu hore 50 ahaa, CursorBench-na lead fiican ayuu ku qaatay.
Qiimaha API-ga, Opus-kii hore input-ka wuxuu ahaa $5 halkii million token, output-kuna $25. Opus 5.5 input-ku waa $4, output-kuna waa $20. Anthropic waxay bilowday inay xalliso efficiency-ga. Kadib waxaan ku tijaabiyay app-ka Claude: waxaan ka codsaday inuu ii dhiso application home automation ah ee aan codkayga ku maamulo nalalka iyo albaabada guriga. Su'aalihii uu u baahnaa ayuu i weydiiyay, kadibna qiyaastii 30 daqiiqo ayuu ku dhammeeyay. Waa Next.js application, codka waxaa loo isticmaalay speech recognition-ka ku dhex jira Chrome, amarradana waxaa kala saara Jev oo aan OpenRouter ku isticmaalay.
Markaan dhaho "turn off all the lights" ama "close the living room door", isla markiiba wuu fuliyaa. Usage-kayga toddobaadkan wuxuu ahaa 9%, markaan dhammeeyay application-kana wuxuu noqday 10%, macnaha 1% oo keliya ayuu isticmaalay. Aniga waxaan u arkaa model promising ah, marar badanna waan isticmaali doonaa. Application-ka GitHub ayaan soo gelinayaa, link-iguna wuxuu ku jiri doonaa description-ka. Comments-ka iiga sheega sida aad u aragteen haddaad tijaabiseen.
English summary
A six-minute Somali-language (Af-Soomaali) AI news and demo video from Dugsiiye on Claude Opus 5.5, Anthropic's new model. Mohamed covers Anthropic's claim that it matches Fable on most work at 40% lower cost than the previous Opus, and why token usage has been the weak point of Opus models. He walks through the benchmarks: 66.4 on agentic coding against Fable's 55, 54 on Terminal-Bench, up from 50, and a clear lead on CursorBench. API pricing drops to $4 per million input tokens and $20 per million output. He then has Claude build a voice-controlled home-automation Next.js app in about 30 minutes, using Chrome's built-in speech recognition and Jev through OpenRouter, and demos it switching lights and closing doors. The whole build used only about 1% of his weekly usage. Free on YouTube.
Waxaad ku baranaysaa
- Waa maxay Opus 5.5 iyo waxa uu ka duwan yahay models-kii Opus ee hore
- Sababta token usage-ku u ahaa dhibaatada ugu weyn ee models-ka Opus
- Natiijooyinka benchmark-yada agentic coding-ka, Terminal-Bench iyo CursorBench marka la barbardhigo Fable
- Qiimaha API-ga: input-ka iyo output-ka halkii million token
- Sida Claude u dhisay application home automation ah ee Next.js ku dhisan
- Isticmaalka speech recognition-ka Chrome iyo Jev si codka loogu maamulo nalalka iyo albaabada
- Inta usage ah ee dhismaha application-ku qaatay
Cutubyada
Su'aalo
- Maxaa Opus 5.5 ka duwan Opus-kii hore?
- Anthropic waxay sheegtay inuu shaqada inteeda badan kula siman yahay Fable, isla markaana uu 40% ka jaban yahay Opus-kii hore. Sidoo kale token-nada si ka wanaagsan ayuu u isticmaalaa, dabeecaddii Opus ee dadku jeclaayeenna waa lagu soo celiyay.
- Immisa ayuu ku kacayaa?
- API-ga, input-ku waa $4 halkii million token, output-kuna waa $20. Opus-kii hore wuxuu ahaa $5 input-ka iyo $25 output-ka.
- Sidee loo dhisay demo-ga home automation?
- Waxaan app-ka Claude ka codsaday inuu dhiso Next.js application aan codka ku maamulo nalalka iyo albaabada. Codka waxaa qaata speech recognition-ka Chrome, amarradana Jev ayaa kala saara isagoo OpenRouter la adeegsanayo. Qiyaastii 30 daqiiqo ayuu ku dhammeeyay, link-ka GitHub-na waxaad ka heli kartaa description-ka.
