BellingBird

The world is an Open Source.
Make it count
#Nextbigthing

[BellingBird @BellingBird]

Our reason makes us see beyond life and time; our passion and sentiments blinds us again and again.

[BellingBird @BellingBird]

Deepseek saves the world again.

[shirish @shiri_shh]

running Kimi K3 locally is easy. you just need:

- 1.5 TB of GPU memory
- roughly 8 H100s
- a small power substation

[BellingBird @BellingBird]

I've got 3090 machine.

[BellingBird @BellingBird]

@Kimi_Moonshot Kimi actually meant what they said by open weight.. like full weights with the whole technical report for closed model to distill 😁

[BellingBird @BellingBird]

Kimi K3's weights are now officially open. A massive step for open-weight models.

[BellingBird @BellingBird]

@Kimi_Moonshot Kimi actually meant what they said by open weight.. like full weights with the whole technical report for closed model to distill 😁

[BellingBird @BellingBird]

@Kimi_Moonshot Big respect to Moonshot AI for dropping not just the weights, but the full technical report and even the infrastructure stack behind it. This is the kind of move that accelerates the entire open-source ecosystem.

[BellingBird @BellingBird]

Kimi K3's weights are now officially open. A massive step for open-weight models.

[Kimi.ai @Kimi_Moonshot]

Releasing the model weights and technical report of Kimi K3.

Kimi K3 is our most capable model: a 2.8T MoE model with native visual understanding and a 1M-token context window.

New model architecture: 2.5x the intelligence per unit of compute, not just more params.

Alongside Kimi K3, we're opening up more of the stack behind it — high-performance attention kernels, MoE communication library, and infrastructure for running agent environments at scale.

Model weights: https://t.co/7m7eEg6Y0B
Tech report: https://t.co/yeu6cjpMCT
Tech blog: https://t.co/YTfiMSNM1f

[BellingBird @BellingBird]

Kimi K3 is getting hammered with demand.

Free users are being locked out everywhere, and most people can't even try it without paying.

Tomorrow, Zyloo is offering 7 hrs of free access.

Just visit https://t.co/MdOHjtBkZ8 when the event goes live, it starts at 11:00AM NY time.

[BellingBird @BellingBird]

Kimi K3 is getting hammered with demand.

Free users are being locked out everywhere, and most people can't even try it without paying.

Tomorrow, Zyloo is offering 7 hrs of free access.

Just visit https://t.co/MdOHjtBkZ8 when the event goes live, it starts at 11:00AM NY time.

[BellingBird @BellingBird]

Deepseek, Kimi is a distill is what they keep saying.

Meanwhile Cursor got caught lifting Kimi‑K2.5's entire weight

Grok scraped user data without permission.

OpenAI had their hands in Apple's data.

Anthropic got caught doing the same thing.

But Chinese model is the issue.

[BellingBird @BellingBird]

the app gets you one draw.

Just use your referral link on alternate accounts to stack draws.

Each one gives you a random membership length, so you might get 3 days or you might land a full year. And running a few accounts adds up to a solid stretch of free access..

[BellingBird @BellingBird]

Kimi K3 Just Dropped and It's a Beast.

It's been topping leaderboards and even outperforming Claude Fable 5 and Claude Opus 4.8 in coding and agentic benchmarks.

The wild part is you can actually use it for free anywhere from 3 days up to a full year. All you have to do is

[BellingBird @BellingBird]

signup:

https://t.co/52UMaszfOp

Register with your Gmail or mobile number, download the Kimi Work app, and log in.

Once you're in, go to the bottom left corner and click on Invite to Earn to grab your own referral link.

Every person who signs up through you and downloads

[BellingBird @BellingBird]

Has anyone ever actually clicked a Gmail sponsored ads usually message..

been using it for years not once have I clicked even by mistake.

[BellingBird @BellingBird]

@levantolabs I'll be checking out the preview, if this works as advertised, you're solving the exact headache that keeps me from trusting most agents with real work.

[BellingBird @BellingBird]

rt didn't know mark has an x account.

[Mark Zuckerberg @finkd]

(1) Today we're releasing Muse Spark 1.1 -- a strong agentic and coding model at a very low price. It's available through our new Meta Model API and in Meta AI.

[BellingBird @BellingBird]

@xiaofen88377379 Aus

[BellingBird @BellingBird]

A Japanese AI lab SakanaAI, just built a model matching Fable and Mythos.

their model, Fugu Ultra, scores: 54.2 on SWE-Pro, 95.1 on GPQA-D, 93.2 on LiveCodeBench v6. beats Opus, Gemini 3.1, and GPT 5.4 across all.

and offers zero export control risk.

https://t.co/zFJHcvLK1m

[BellingBird @BellingBird]

I thought you knew.