Jim Bohnslav (@jbohnslav) 's Twitter Profile
Jim Bohnslav

@jbohnslav

computer vision + machine learning. Perception at Zoox. Prev: Cobot, PhD. Arxiv every day.

ID: 246171448

linkhttps://jbohnslav.github.io/ calendar_today02-02-2011 08:33:55

6,6K Tweet

1,1K Followers

4,4K Following

Jim Bohnslav (@jbohnslav) 's Twitter Profile Photo

Coding agents write code so fast that PRs are huge. I don't have time to read it all carefully. Now I need a coding agent just for review 😑 the agents are perpetuating themselves

Prime Intellect (@primeintellect) 's Twitter Profile Photo

Introducing the Environments Hub RL environments are the key bottleneck to the next wave of AI progress, but big labs are locking them down We built a community platform for crowdsourcing open environments, so anyone can contribute to open-source AGI

Andi Marafioti (@andimarafioti) 's Twitter Profile Photo

Fuck it. Today, we open source FineVision: the finest curation of datasets for VLMs, over 200 sources! > 20% improvement across 10 benchmarks > 17M unique images > 10B answer tokens > New capabilities: GUI navigation, pointing, counting FineVision 10x’s open-source VLMs.

Fuck it. Today, we open source FineVision: the finest curation of datasets for VLMs, over 200 sources!

> 20% improvement across 10 benchmarks
> 17M unique images
> 10B answer tokens
> New capabilities: GUI navigation, pointing, counting

FineVision 10x’s open-source VLMs.
Jim Bohnslav (@jbohnslav) 's Twitter Profile Photo

openAI hit a critical threshold with Codex: even if it's not quite as good as claude code, it's close enough that I'd rather only spend the $200 at OpenAI.

vik (@vikhyatk) 's Twitter Profile Photo

Excited to release a preview of Moondream 3. A 9B param, 2B active MoE vision language model that makes no compromises; offering state-of-the-art visual reasoning while still retaining an efficient and deployment-friendly form factor.

Jim Bohnslav (@jbohnslav) 's Twitter Profile Photo

you know you're having a bad day when you start putting shit like this at the top of your code torch.backends.cuda.matmul.allow_fp16_reduced_precision_reduction = False torch.backends.cuda.matmul.fp32_precision = "ieee"

Junyang Lin (@justinlin610) 's Twitter Profile Photo

This is the 5th shot! Super crazy! We opensourced a 235B-A22B Instruct and Thinking Qwen3-VL models under Apache 2.0! Qwen3-VL, the new generation of our vision-language model, whose previous version was released a long time ago. During these days, we have conducted a lot of

Jim Bohnslav (@jbohnslav) 's Twitter Profile Photo

qwen3-vl > ... we shift from the conventional... practice of injecting visual tokens into only a single layer, to injecting them across multiple layers of the large language model

qwen3-vl

> ... we shift from the conventional... practice of injecting visual tokens into only a single layer, to injecting them across multiple layers of the large language model
Jim Bohnslav (@jbohnslav) 's Twitter Profile Photo

Qwen3-VL-235B-A22B is a beast. I'd give you my vibe check, but at 471GB, it's still downloading 😅 vLLM and 🤗 transformers reference a 30B-A3B MoE and 32B, 4B dense models. I expect almost everyone will use these dramatically smaller sizes. Look forward to trying them out!

Qwen3-VL-235B-A22B is a beast. I'd give you my vibe check, but at 471GB, it's still downloading 😅

vLLM and 🤗 transformers reference a 30B-A3B MoE and 32B, 4B dense models. I expect almost everyone will use these dramatically smaller sizes. Look forward to trying them out!