Hayabusa

METÜ TIP

[Hayabusa @EbuHayabusaa]

@AlicanKiraz0 maşallah abi hayırlı olsun

[Hayabusa @EbuHayabusaa]

@AlicanKiraz0 kanım dondu

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms https://t.co/chMHTs8qW6

[Alican Kiraz @AlicanKiraz0]

Hi everyone! This time, instead of using a single-shot prompt, I compared the models using a Planning → Development workflow with OpenCode + High Effort. The judge was GPT-6 Astra Pro. 🔥🦾

Models:
- DeepSeek v4.1-1017
- Fable 5.2
- Gemini 4 Carbon
- GPT 6.1-Sol

Winner in terms of quality/cost balance:
- GPT 6.1-Sol — 91/100

For the independent evaluation, I focused on physics accuracy, cargo and flooding behavior, measurement reliability, code structure, and Canvas rendering quality.

Raw Technical Performance and Costs:
- Fable 5.2 — 93/100 — $25.80
- GPT 6.1-Sol — 91/100 — $1.68
- DeepSeek v4.1-1017 — 80/100 — $0.692
- Gemini 4 Carbon — 76/100 — $1.26

(Performance + Quality) / Price Scores:
- DeepSeek v4.1 Flash 1017: 100/100
- GPT Sol 6.1: 52.2/100
- Gemini 4 Carbon: 46.9/100
- Fable 5.2: 3.1/100

Comparisons:
- Most comprehensive simulation and flooding model: Fable 5.2
- Best cargo/collision engineering: GPT 6.1-Sol
- Best cinematic ship presentation: Fable 5.2
- Best ocean and storm atmosphere: DeepSeek
- Best quality/cost balance: GPT 6.1-Sol

Fable's 2-point advantage in this evaluation comes at approximately 15.4× the cost of GPT 6.1-Sol.

[Hayabusa @EbuHayabusaa]

@marketingwai @Avenoxai Reis soylermisin rica etsem

[Hayabusa @EbuHayabusaa]

@st_nerona @Avenoxai Reis hala çalışıyorsa rica etsem soylermisin

[Hayabusa @EbuHayabusaa]

@AlicanKiraz0 @onandagpt Bizide rtx sparkla gör abi

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms Haha

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms Argon 4ü kullandim

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms Reis insana hakeret etmedimki carbona salak dedim

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms tamam anlarsın salak olduğunu

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms yarma ya şimdi sen carbon ,fabledan opusdan daha mı iyi diyorsun

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms https://t.co/To5RMAWwR9 sen şunun gibi bişeyi tek promtla gemini modellerine yaptırabilirmisin

[Felix Rieseberg @felixrieseberg]

I wanted to test Fable 5.2's quality with one-shot prompts without human feedback.

Everything in this game is from scratch: models, sounds, textures. ~$550 at API pricing, ~2.7M output and 445M input tokens, heavily cached. Took 8 hours!

Play it here: https://t.co/fazOCcyNIH

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms sen hiç opus 5.5 yada fable kullandınmı

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms hahahh

[Hayabusa @EbuHayabusaa]

@mubarek_adam @ylmzzms inanmam malın teki googleun modelleri carbonda buna dahil

[Hayabusa @EbuHayabusaa]

@sophiachan80 hemşireliğide bitiremez

[Hayabusa @EbuHayabusaa]

@beklersiker @ylmzzms gine geçemez bu googledan bişey olmaz

[Hayabusa @EbuHayabusaa]

@beklersiker @ylmzzms güzel dostum bu arena gemini app de mi yapıyor benchmarkı güzel kardeşim benim

[Hayabusa @EbuHayabusaa]

@beklersiker @ylmzzms arenada testler yapılıyor zaten ve her an canlı testlerde devam ediyor benchmarklar kim takar antigravityi