Fisherman

Here just to support good open source, Linux Gaming and AI projects.

[Jun Song @jun_song]

Yes. If you don’t understand why people are jumping into Local AI, you are not smart.

[Jun Song @jun_song]

Kimi-K3 feels like Opus-5 with superfast mode enabled.

[Migel Tissera @migtissera]

I'm releasing Tess-4-9B.

Tess-4-9B is the compact sibling of Tess-4-27B, built on the Qwen/Qwen3.5-9B-Base model.

It was post-trained on the same deliberate blend as the 27B release: 64K-token long-context agentic traces — real engineering work done with Fable-5, not synthetic generations — with a reasoning style approximated from Fable-5 by a three-model teacher ensemble (Opus-4.8, GPT-5.5, and GLM-5.2) fused into one coherent voice.

Significant improvement on all measured axes against Qwythos-9B.

[Pasha @pashakho]

We grew a very big model in a very tiny pot. 🌱

Meet Bonsai 27B; a 27B-class multimodal model that fits on a phone.

🪴 3.9 GB at 1-bit
💻 5.9 GB in ternary
🧠 Reasoning, vision, tool use
🔓 Apache 2.0

Small footprint. Full-grown intelligence.
https://t.co/3QFvwYdGcj

[Georgi Gerganov @ggerganov]

the 10000th release of llama.cpp 🥳

[0xSero @0xSero]

8gb VRAM best model for your hardware

1 bit gguf doing better than models 25x its size!?!?! This will even run in 4gb (:

https://t.co/jyNyvOb6fg

[Null-O @GOROman]

ブラウン管（CRT）のエミュレーション作ってるんだが、スーパーリアルになったw

[Fisherman @Rybens92]

@PrismML I don't know if you all understand how revolutionary it is!

This density of intelligence, so you can run it even on 8 gb VRAM with reasonable context window if kv cache size allows this...

This is truly "Free intelligence for everyone"!

[PrismML @PrismML]

Today, we’re announcing Bonsai 27B: the first 27B-class model to run on a phone.

Bonsai 27B is the new multimodal flagship of the Bonsai family. Based on Qwen3.6 27B, it brings a new capability tier to local AI: multi-step reasoning, structured tool use, long-context workflows, and coherent agentic loops.

Until now, models in this class have been impractical to deploy locally. A 27B model occupies roughly 54 GB in 16-bit precision, and even a strong 4-bit build is around 18GB - too large for a phone and for most laptops.
Bonsai 27B changes that.

It comes in two variants:
• Ternary Bonsai 27B: 5.9 GB, 1.71 effective bits per weight, optimized for laptop-class quality.
• 1-bit Bonsai 27B: 3.9 GB, 1.125 effective bits per weight, optimized for phone-class footprint.

Everything is open-sourced today under the Apache 2.0 license.

[Tengfei Wang @DylanTFWang]

⚡️ Tencent HY-World 2.1 is HERE!

Not a video. A world. 🪐

3 months just after 2.0, we've supercharged everything:
✨ Cleaner geometry
✨ Sharper rendering
✨ Larger explorable range

It's a REAL 3D world you can Walk through and Interact with.

[Lisan al Gaib @scaling01]

The Sonnet tier of models was already dead before, with Kimi-K3 I expect the Opus tier to become obsolete too

Frontier Labs have to go to 10T models now

[leo 🐾 @synthwavedd]

well hello there

[Tony Simons @tonysimons_]

@claudeai Careful teachers: they’ll just take it away from you in a week and then give you back a nerfed version after the USG slobbers all over it.

[Fisherman @Rybens92]

@LottoLabs That's the direction I want labs to go almost all in!

1bit quant with performance almost on par with 8bit is so sexy 😄

[Lotto @LottoLabs]

Very interesting to see more attention on 1b quants

Idk how well it will actually benchmark but once we get decent models at 1bit that changes the game

[BridgeMind @bridgemindai]

Fable 5 just DOMINATED GPT 5.6 Sol.

And this time there is no debating it.

BridgeBench V3, our new arena, runs both models on the identical task and a judge panel picks the winner.

Final score: Fable 5 wins 67-10. An 87% win rate.

Scoreboards said they were neck and neck.

Head to head, Fable 5 is destroying GPT 5.6.

BridgeBench V3 alpha is open source.

Link below. See every match yourself.

https://t.co/2xmbOGFWiL

[Ivan Fioravanti ᯅ @ivanfioravanti]

Now you understand why paying more for Local AI does make sense. @TheAhmadOsman has always been right since the beginning.

[Ornith @ornith_]

Thanks for all the love and support❤️❤️❤️

🐦Ornith will be open source and will continue pushing local LLM experience forward.

🫡Please tell us if you have any issue or suggestion.

[Ivan Fioravanti ᯅ @ivanfioravanti]

K2.7 Code +136% with DSpark! Thanks Novita AI 🙏

[Tony Simons @tonysimons_]

Massive Hermex update just dropped!

If you’re on iOS and not already Hermex-ing it up, this is currently the best way to access and control your Hermes Agent from your iPhone — and now iPad!!

@uzairansar cooked on this one! 🔥🔥🫡