Paka Joe retweeted
Para quem quer ver alternativas ao grok Bot, coloquei todasminhas anotações e analises que já fiz no Blog:
my-tierroom.party/blog
Tem mais de 80 avaliações.
Alguns que estão lá:
@mausbot_ @OpenBot_ @agentes_chat @rakazobot @tucanosh @ai_timon Anymodelbot @NousResearch @lbki34064963 @XXXX540515 @olekss @letta_ai @MeuZe_bot @pdp @NuphosAI @CopilotKit and much much more.
São muitos modelos, para todos os gostos. E vou atualizando semanalmente o que eu revisar novamente, a depender do que vem atualizaddo. Aproveite para votar no seu preferido.
Minha recomendação "facilidade" de olho fechado:
Mausbot, OpenBot, Codync, Bloks, Tucano, Noodle, Nuphos, Rakazo, Timon, Zé Bot..
Teams, ainda o Agentes.chat está na frente para mim, mas estão chegando, lançando ferramentas de times.
Paka Joe retweeted
I spent the past ~month working with this enormous dataset of real-world coding agent transcripts from @FreebuffHQ! It is anywhere from 31x to 241x bigger than (the expanded) SWE-chat, at 251B total tokens. It's so large that running this dataset through Sol 6.1 once costs $500k.
Paka Joe retweeted
QoderCN 的deepseek 4.1 flash 也被实锤会路由到qwen3.8系列
速度明显过慢,而且ai分析指纹偏向qwen3.8家族
白嫖就好,一言不和就排队,果然掺水了,别付费
Paka Joe retweeted
$52M seed round. 15M+ users. But this is what we're most proud of yet.
Introducing Drama 3: the most emotive and controllable TTS model in the world
And yes, it's open-weight.
You can now control how even a single word should sound, change its tone, and shift your voice mid-sentence.
Uncle Roger was impressed too.
Try it here:
s.fish.audio/rqh0l8
Comment “DRAMA” within the next 48 hours for 1 month free.
If you're an enterprise your first month is on us:
s.fish.audio/qz9r0t
Paka Joe retweeted
Step 5 Preview is now live on OpenRouter
I gave 4 models the same 3 tasks: build an Eiffel Tower, a rocket launch, and a building collapse all in Three.js
Step 5 Preview came out cheapest and fastest across the board and got good results not expecting this from an open-weight model at all. DeepSeek V4 close behind. GLM 5.3 and Qwen 3.8 took longer and cost more, but still landed solid results.
Video says more than I can.
Paka Joe retweeted
Did you know…
Jesus Christ was originally going to be named Gary until Mary stubbed her toe.
Paka Joe retweeted
Love this fucking printer so much. It just works without any bullshit or gay app, absolute champ. Have owned like 3 of them and will never buy any other printer again until I die
Paka Joe retweeted
Xiaomi MiMo-V2.6-Pro is #6 on World of AI Bench that scores real shipped work, not homework tests.
Composite 83.3. That’s 0.4 behind Claude Opus 5 and a tick ahead of GPT-5.6 Sol.
MiMo-V2.6-Flash is #17 at 79.0.
Open weights. Same price as V2.5. Already in the frontier cluster for frontend, game dev, 3D, agents, and creative code.
woaibench.ai/
Paka Joe retweeted
I tested Xiaomi MiMo-V2.6 Pro on my LLM benchmark and it is the highest scoring open-weight model seen so far. Beats Muse Spark 1.3 on price.
👉 bench.killswitch-lang.org/
Paka Joe retweeted
xiaomi mimo v2.6 pro just refactored my entire codebase in one call
67 tool invocations. 1M+ new lines. 24 brand new files
it modularized everything. broke up monoliths. cleaned up spaghetti
none of it worked
but boy was it beautiful
Paka Joe retweeted
Tried same test with MiMo v2.6 Pro (Pelican svg test)
Do you think the output is better than Grok 4.7?
Tried it in OpenCode
cost: $0.05
time: 6m 9s
Grok 4.7 what are you doing??? 💀
it’s supposed to be on par or better than opus 5
what is this output ?
left - grok 4.7
right - opus 5
cost $1.2 (72k tokens)
worked for 14 minutes
conclusion: it failed the pelican svg test
grok 4.8 next teapawt.org/dtzy_88/status/2102071…































