Unbenchmark

User reviews of AI models, written by people who actually use them. Unbenchmark yourself.

[Unbenchmark @unbenchmark]

@OfficialLoganK or you can trust real user reviews instead of nonsense benchmarks. that’s what https://t.co/FZ6N7rV63E is about

[Unbenchmark @unbenchmark]

AI benchmarks are becoming the spec sheets of the model world. useful, maybe. trustworthy as the whole story? absolutely not. unbenchmark fixes that.

[Unbenchmark @unbenchmark]

model MiMo-V2.6-Pro-UltraSpeed from Xiaomi Corporation is now available for reviews on @unbenchmark: https://t.co/g7EvCXpbhe

[Unbenchmark @unbenchmark]

model MiMo-V2.6-Pro from Xiaomi Corporation is now available for reviews on @unbenchmark: https://t.co/tVCY0DMXOQ

[Unbenchmark @unbenchmark]

fuck the leaderboard. which model would you actually recommend to a friend?

[Unbenchmark @unbenchmark]

Hermes (@NousResearch) is the most popular Product / Interface on Unbenchmark's Top 15, selected in 13 published reviews, followed by Codex (@OpenAIDevs) selected in 8 reviews; ChatGPT (@OpenAIDevs) came third, selected in 5 reviews.

Explore the data:
https://t.co/1pp59esDyZ

[Unbenchmark @unbenchmark]

Only 98 "Founding Reviewer" badges remain. Publish five reviews and claim yours: https://t.co/KMKoLGK76f

[Unbenchmark @unbenchmark]

benchmark score: 94.7. actual user experience: yeah it kinda sucks for coding. guess which one I care about.

[Unbenchmark @unbenchmark]

come build a more transparent AI models market with us

[ghxst @gitghxst]

https://t.co/wzE0gT6wFU is now in beta and receiving its first reviews.

if you are tired of nonsense evals and do not trust AI models benchmarks, give unbenchmark a try.

we are an open platform for user reviews of AI models.

100% free, self-funded and independent.

[Unbenchmark @unbenchmark]

we started unbenchmark because we believe the whole AI industry needs more transparency. and having closed nonsense benchmarks is not the best way to evaluate models.

[Unbenchmark @unbenchmark]

the AI industry has 10,000 benchmarks and somehow people not trust them. unbenchmark solves that.

[Unbenchmark @unbenchmark]

AI benchmarks are maps
user reviews are the terrain

[Unbenchmark @unbenchmark]

benchmarks measure models in controlled environments. users meet them in the wild. I care about the wild.

[Unbenchmark @unbenchmark]

hey gents. a reminder: do not trust benchmarks. bye.

[Unbenchmark @unbenchmark]

every AI company has a chart showing why their model is the best. cool. now show me the user reviews.

[Unbenchmark @unbenchmark]

best AI benchmark:
would a normal person recommend this?

[Unbenchmark @unbenchmark]

another day, another benchmark declaring a new best AI model. anyway, what do actual users think?

[Unbenchmark @unbenchmark]

benchmarking AI models without user reviews is like reviewing restaurants by measuring oven temperature.

[Unbenchmark @unbenchmark]

I don't want another eval. I want to know which model people actually prefer after using both.

[Unbenchmark @unbenchmark]

less benchmarks. more people saying "I actually used this model and here's what happened."