Sabitlenmiş Tweet
m0h
58.2K posts
m0h
@exploraX_
how to AI | check: https://t.co/iToThATakz
Katılım Eylül 2019
504 Takip Edilen25.6K Takipçiler
the second chinese 2T-parameter model (Qwen 3.8) is now live for preview and it's FREE for everyone to try.
alibaba promised that this model the second only fable 5, meaning that the open-weight model will also be a frontier model like it brother kimi k3.
btw, there are already rumors that the U.S. government could ban Chinese AI models in the United States.
if that happens, the release of Qwen 3.8 may only accelerate the debate.
try Qwen 3.8 here: chat.qwen.ai
Qwen@Alibaba_Qwen
Qwen3.8 is launching and going open-weight soon!🌐 With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5. You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out. Can't wait to hear what you build. Stay tuned! 🚀 Token Plan international：qwencloud.com/pricing/token-… China：platform.qianwenai.com/pricing/token-…
English
Alibaba just made their 2.4T param flagship model for free
qwen 3.8 max preview, no benchmarks published yet but it's live right now at chat.qwen.ai
i think its fable 5 and gpt 5.6 sol level model and surpassed kimi k3
what you get for $0:
> 2.4T parameters, alibaba's biggest model yet
> web dev mode built in, generates full sites from a prompt
> no api key, no credit card, just open the site and pick the model
> positioned by alibaba as second only to claude fable 5, unverified but that's their claim
what this replaces:
> waiting on paid api access to test frontier chinese models
> burning credits just to see if a new model is worth switching to
how to try it (10 seconds):
chat.qwen.ai → select "qwen3.8-max-preview" → toggle "web dev" → prompt
important: still in preview, alibaba says full open weight release is coming but no date yet. benchmarks aren't published so treat any numbers you see elsewhere as unconfirmed
test it while it's free and see for yourself
Isra@israfill
FREE MiniMax M3 + all models API key - fresh key, full 5B tokens in it 😳 yes im dropping another one, not the leftover key with brand new quota. ive already given away 10B+ tokens across these MiniMax keys. people actually burn through them and M3 still holds up against Claude Sonnet on my coding runs with 1M context. heres what this key unlocks: - MiniMax-M3 (frontier coding, 1M ctx, competes with Claude Sonnet) - MiniMax-M2.7 / highspeed (general + turbo) - MiniMax-VL-01 (vision-language) - Hailuo 2.3 / 2.3Fast (video gen) - speech-2.8-hd + turbo (TTS) - Image-01 (image gen) - Music-2.6 + Cover (music gen) coding, video, voice, image, music. one key. you would pay ~$330/mo stacking claude + elevenlabs + midjourney + runway + suno for the same jobs API Key: sk-cp-_NMLvVX-xQtJdX8KpzxTlLD4TVJAu_B_HlV_7j73-6J9-v131owHmNYR1DNxlurP_fcrH72tJKjFmFct3VDR4eRicDKe-AsJwvMYdP1yeIFXVKh7aDT0uik > paste into any OpenAI-compatible client (base URL in reply) > shared key so it drains faster the more people hop on use this before someone empties it
English
@VuAnh20091987 yes yes, deepseek v4 pro has a better overall performance than fable 5 but not forgetting that I didn’t give both the same prompt rather I dropped the video to fable 5 and told it to recreate it
English
@exploraX_ Polish vs mechanics is the ultimate LLM tradeoff. Polish means better UI instruction following, but raw mechanics show actual spatial logic
English
DeepSeek V4 Pro VS Fable 5 🔥
I uploaded the game video to my claude code agent and I simply told fable 5 to recreate the game exactly.
the result? honestly closer than I expected.
> both models nailed accuracy with targeting at varying distances.
> both built the full gunsmith: barrel, muzzle, optic, stock, mag, finish + live weapon stats.
> both got the HUD right: score, hits, accuracy, fire mode, ammo.
where they split:
> Fable 5 came out more polished. enclosed range with proper booths, labeled ammo crates, denser target field, and hit popups that show distance + headshot/down. The gunsmith even explains each part's tradeoff.
> DeepSeek V4 Pro held its own with an open outdoor range, skybox, and one thing it showed more clearly: attachments actually changing the gun model on screen.
Verdict: Fable 5 wins on polish and detail. DeepSeek nailed the harder mechanic.
I've link the fable 5 game below.
Chetaslua@chetaslua
🚨 Deepseek V4 pro < api is sus> i have read multiple reports on x that it is routing to fable 5 , if it is right then we can see the jump in performance normal If it is not routing to fable 5 < quite impossible > then they made better model than kimi k3 , here is output
English
my app just hit 55 users
made Pane, a free and open source app to track all your AI subs limits in one place
haven’t even pushed it much yet, but it’s already averaging 20+ daily active users
thanks for all the support fr <3
J A Z I I@notjazii
Same prompt but different outputs most people judge models by their frontend skills, so i decided to test: > Kimi K3 > GPT 5.6 Terra > MiniMax M3 > GLM 5.2 all four were tested on same reasoning with same prompt which one built the best frontend?
English
@exploraX_ all hail Moh !!
all hail Moh !!
all hail Moh !!
feeding to my agents now!!!
English
wikipedia released a document that lists everything that make a writing feel AI.
it's called Signs of AI Writing, and you can use it to train your AI to write content that feels genuinely human.
i've converted the document into a skill, and here's how you can easily install it in 2mins:
> the whole thing is one file. drop it in your skills folder and your agent picks it up automatically on any writing task.
> one command does it:
git clone github.com/Moh4696/be-hum… ~/.claude/skills/be-human
that's it.
the repo is in the comment section.
m0h@exploraX_
google's notebookLM charges $19.99/mo for pro. this open-source alternative gives you unlimited access for $0. zero subscription. runs locally. MIT licensed. features: → podcast generation → multi-modal ingestion → up to 6 languages → universal LLM selection github repo link in the comments.
English
Kimi is suffering from success
they’re pausing new subscriptions cuz demand pushed their compute close to its limits
understandable when it’s already the third best model in the world
already got APIs from team, so i’ll keep testing
did they actually hit capacity or is this just genius marketing?
Kimi.ai@Kimi_Moonshot
Kimi K3 has received far more love than we expected, and our GPUs are feeling it. Over the past 48 hours, demand has pushed close to the limits of our current capacity. To protect the experience of existing subscribers, we're temporarily pausing new subscriptions and prioritizing compute for current members. Existing subscribed users are not affected. We're adding capacity as fast as we can and will reopen new subscription spots in batches. Going forward, we'll also split membership into two more focused plans: Kimi Membership for Kimi Web, App, and Work; and Kimi Code Membership for coding workflows. This will help us match compute more precisely and keep the experience stable. Thank you for your patience and understanding!
English
