ByteShape

Accelerating AI by Learning Optimal Datatypes

[ByteShape @ByteShape]

We released Qwen 3.6 35B A3B GGUF quants in both NTP and MTP.

The benchmark results made one thing clear: size, speed, and quality do not move in a straight line.

GPU-5 was hard to beat. If it fits, try it first.

Blog: https://t.co/WJrOEKPjfk

[ByteShape @ByteShape]

@Capetlevrai @0xSero 👀

[ByteShape @ByteShape]

@KC_goes_digital We have them on our radar.

Our models are optimized for the best quality vs speed/size tradeoff. We benchmark on real tasks and measure performance across different hardware, so it’s easier to find what actually works best on your setup.

[ByteShape @ByteShape]

We recently released our Qwen 3.5 35B A3B quants.

If your setup can run GPU-7, you should try it.
If not, we’ve got options across all hardware.
Pi → 5090.

Blog: https://t.co/d1TCjbIgMS

[ByteShape @ByteShape]

@therealsol4ra Regardless, we're happy for everyone to perform independent evaluations of our models and test them out for yourselves! We're highly confident in their quality!

[ByteShape @ByteShape]

@therealsol4ra KLD is a distribution deviation metric measuring in this case the deviation in token generation between quants and the original model. However, it does not measure behaviour, like the model taking a different path to solve the same problem than the original model for example.

[ByteShape @ByteShape]

@Yeely1310 @ravikiran_dev7 We're glad our quants are not falling for that👀

[wd 🔺 @populartourist]

+ Hermes Agent
+ Qwen3.5 35B A3B
+ 4x parallel agents with 262k context window (each)
+ Over 200 t/s token generation + 3000 t/s prefill
+ 23.2GB total VRAM consumption on RTX 5090

It can take 5 parallel agents, 4 was the sweet spot with 2x in completion time vs 1.74x.

Dream inference.

@NousResearch
@ByteShape

[ByteShape @ByteShape]

Run your own local AI coding agent

We just published a beginner guide for using @opencode with local models (@lmstudio, llama.cpp, @ollama).
Mac, Linux, WSL2, full setup + API + config.

https://t.co/TzgNNTxU49

From “I have a model” → “I have a working coding agent”

[ByteShape @ByteShape]

GPUs are consistent. CPUs are not.

With our ByteShape Qwen 3.5 9B quants, the same models perform well across GPUs, but CPUs each have their own “favorites”.

No one-size-fits-all. Optimize for your hardware.

https://t.co/RSX3iK3vgh

[Allen Lau 🇨🇦 @allenlau]

ByteShape was quietly launched just before the year end. Two weeks ago, we announced our investment in the company. Since its launch, and with minimal fanfare on purpose, @ByteShape cumulative downloads have easily blown past 100,000. No small feat for a new startup!

[Allen Lau 🇨🇦 @allenlau]

Announcing @twosmallfishvc's investment in @ByteShape.

In short, ByteShape is delivering step-function gains in AI efficiency, including up to 7x faster training, up to 10x faster inference, plus up to 40% compression to reduce model size.

[ByteShape @ByteShape]

@atarl666028 @Alibaba_Qwen 👀

[wd 🔺 @populartourist]

Excellent quantisation on Devstral Small 2

[ByteShape @ByteShape]

We released ShapeLearn-optimized GGUFs for:

• Devstral Small 2 24B, tuned for RTX 40/50 GPUs
• Qwen3 Coder 30B, runs everywhere, yes even the Pi

Maximum quality. Fastest TPS. Minimal compromise.

GGUFs + interactive plots are live: https://t.co/VVZ87Pvm1p

[ByteShape @ByteShape]

Edge computing is getting spicy! Shoutout to @geerlingguy for showcasing our model. Love seeing what the community is building and how hard it’s being pushed. Clip: https://t.co/aPvRpsxIAC

[Jeff Geerling @geerlingguy]

Raspberry Pi has a new AI HAT. This time with built-in 8 GB of RAM, so you can run machine vision + LLM inference all without touching the Pi's CPU. It's $130 and a little bit of a niche item. Find out why in my video: https://t.co/vMhZ5w1wCU

[Jeff Geerling @geerlingguy]

Raspberry Pi has a new AI HAT. This time with built-in 8 GB of RAM, so you can run machine vision + LLM inference all without touching the Pi's CPU. It's $130 and a little bit of a niche item. Find out why in my video: https://t.co/vMhZ5w1wCU

[HackerNewsTop5 @hackernewstop5]

A 30B Qwen Model Walks into a Raspberry Pi and Runs in Real Time #HackerNews
https://t.co/5v7VyAe7ES