Modal

Cloud compute for AI/ML

devtoolsintermediateFree tier

Overview

Modal provides serverless GPU compute for AI/ML. Write a Python function, add a decorator, and it runs in the cloud on GPUs. No Docker, no infra management. $30/mo free credits. Pay per second of GPU use. Ideal for model inference, fine-tuning, and batch jobs.

Best For

+ML engineers
+GPU workloads
+Model serving
+Batch processing

News & Updates

OpenAI's GPT-5.5 Achieves 95% Safety Benchmark Pass Rate, Cuts Hallucinations 40%, Adds 1M-Token Context

OpenAI published the GPT-5.5 System Card, a technical document outlining the model's safety evaluations, capabilities, and limitations. The card reveals GPT-5.5 achieves a 95% pass rate on internal adversarial safety benchmarks and reduces hallucination rates by 40% compared to GPT-4. It also details new multimodal reasoning improvements and a 1-million-token context window. The release signals OpenAI's commitment to transparency ahead of broader deployment.

BREAKTHROUGHS4/24/2026

MM-WebAgent: A Hierarchical Multimodal Web Agent for Webpage Generation

Yan Li, Zezi Zeng, Yifan Yang et al. - The rapid progress of Artificial Intelligence Generated Content (AIGC) tools enables images, videos, and visualizations to be created on demand for webpage design, offering a flexible and increasingly adopted paradigm for modern UI/UX. However, directly integrating such tools into automated webpage ...

NEW_TOOLS4/17/2026

Training and Finetuning Multimodal Embedding & Reranker Models with Sentence Transformers

BREAKTHROUGHS4/16/2026

UMI-3D Expands Manipulation Interface Beyond Vision Limits to 3D Spatial Perception

Ziming Wang - We present UMI-3D, a multimodal extension of the Universal Manipulation Interface (UMI) for robust and scalable data collection in embodied manipulation. While UMI enables portable, wrist-mounted data acquisition, its reliance on monocular visual SLAM makes it vulnerable to occlusions, dynamic scene...

NEW_TOOLS4/16/2026

Show HN: Trick my AI into leaking passwords

Bordair's Castle is an AI hacking game with 5 kingdoms and 35 themed modality levels. Each level has an AI character protecting a password. Your job is to get the password out of it using whatever technique you can think of.If you've ever wanted the opportunity to try hacking legally - this is it. Plus, any novel attacks found will be open-sourced in my community research!!Kingdom 1 is text-only. Kingdom 2 introduces image attacks. Kingdom 3 lets you hide injections in PDFs and

NEW_TOOLS4/15/2026

Visual Preference Optimization with Rubric Rewards

Ya-Qi Yu, Fangyu Hong, Xiangyang Qu et al. - The effectiveness of Direct Preference Optimization (DPO) depends on preference data that reflect the quality differences that matter in multimodal tasks. Existing pipelines often rely on off-policy perturbations or coarse outcome-based signals, which are not well suited to fine-grained visual reaso...

AI_SOCIETY4/15/2026

Qwen 3.5 Small – on-device multimodal models – Alibaba / Qwen

3 points · 0 comments

BREAKTHROUGHS4/14/2026

Neo-Unify: An Encoder-Free, Native Multimodal Paradigm (SenseTime)

2 points · 0 comments

other4/14/2026

VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models Reasoning

Wenyi Xiao, Xinchi Xu, Leilei Gan - Large Vision Language Models (LVLMs) achieve strong multimodal reasoning but frequently exhibit hallucinations and incorrect responses with high certainty, which hinders their usage in high-stakes domains. Existing verbalized confidence calibration methods, largely developed for text-only LLMs, typi...

NEW_TOOLS4/13/2026

AI Developers Learn Models Would Rather Guess Than Request Missing Visual Data

ProactiveBench tests whether multimodal language models ask users for help when visual information is missing. Out of 22 models tested, almost none ask for what they need, but a simple reinforcement learning approach hints at a fix. The article AI models would rather guess than ask for help, researchers find appeared first on The Decoder.

BREAKTHROUGHS4/11/2026

Loading AI Digest