Front Page

SURL AI News

Technical AI news on models, software, chips, open source, research, and infrastructure.

U.S. Department of Energy Launches 278 AI Science Workflow Projects Linking Agent Frameworks, Models, and National Laboratory HPC

The first group of Genesis Mission projects spans nuclear energy, critical minerals, chip design, and commercial fusion. Research teams will share models, agent frameworks, and high-performance computing resources. OpenAI has separately committed Codex access and API credits to support about 2,000 researchers and two large-scale scientific experiments.

2026-07-23

代理基礎設施

Anthropic Memory Store API Changes Pagination and Path Semantics, Potentially Invalidating Old Cursors and Query Parameters

New rules for Claude Managed Agents memory listing took effect on July 22, including stable ordering, directory-style prefix matching, and restricted query depth. This is more than an SDK header change: existing synchronization programs may encounter 400 errors, missing items, or invalid cursors.

2026-07-23

訓練系統

SLAI T-Rex Achieves 34.22% MFU in Full-Parameter Post-Training of a Trillion-Scale MoE on Ascend SuperPOD

The research team redesigned parallelization strategies, communication scheduling, kernels, and operator fusion for the DeepSeek-V4 family, bringing Ascend training efficiency to 2.93× the open-source baseline. The system was also used to train an operations research model on solver-verified data, though more complete public reproduction of the code, data, and hardware measurements is still needed.

2026-07-23

端側推論

BaseRT Directly Drives M5 Neural Accelerators via Metal 4, Achieving Up to 6.4× Faster Prompt Processing Than llama.cpp

BaseRT has added hand-written dense and MoE GEMM and FlashAttention kernels, offloading matrix operations during the prompt-processing stage to the neural accelerators inside the M5 GPU. The team reports substantial leads over llama.cpp and MLX on the M5 Pro, although the figures currently come solely from the developers’ own single-machine tests.

2026-07-23

Latest Reports

Browse all news

程式代理／推論最佳化

SWE-Pruner Pro Reads Coding Models’ Hidden States Directly, Cutting Agent Tokens by Up to 39%

SWE-Pruner Pro replaces a separate classifier with a lightweight prediction head that uses a coding agent’s internal representations to decide whether to retain each line of tool output. Tests with two open MoE models across four multi-turn benchmarks showed token savings, but the method requires inference-server modifications, and some quality gains have yet to be reproduced.

2026-07-23

具身 AI／機器人

RynnBrain 1.1 Adds Contact Points and Native 3D Grounding to an Embodied Foundation Model, Trains VLA Across Three Robots

Alibaba’s DAMO Academy has released three RynnBrain 1.1 models—2B, 9B, and 122B-A10B—adding contact-point prediction and native 3D grounding. The accompanying RynnBrain-VLA is trained across Unitree G1, Astribot-S1, and Tianji Wuji using a unified action space, although evidence for cross-hardware generalization still comes primarily from the team’s own experiments.

2026-07-23

開發者工具

GitHub Redesigns Copilot Impact Dashboard Around Adoption Stages, but Output Still Does Not Equal Engineering Quality

GitHub has added a Copilot impact dashboard that compares merge speed and output across code-first, agent-first, and multi-agent adoption stages. The new interface helps enterprises assess depth of adoption, but still lacks quality signals such as defects, rework, and long-term maintenance costs.

2026-07-23

AI 基礎設施

Anthropic to Expand Compute Capacity by 2 GW With AMD Helios and Use Claude to Optimize ROCm

Anthropic plans to deploy its first AMD Instinct MI450-series GPUs starting in the first half of 2027, with a total capacity of up to 2 GW. The partnership goes beyond hardware procurement: the companies will also use Claude to optimize GPU workloads and the ROCm software stack.

2026-07-23

開放權重模型

Thinking Machines Releases Inkling: 975B MoE Open-Weight Model Supports One Million Tokens and Native Multimodality

Inkling enters the open-weight market with a Mixture-of-Experts architecture featuring 975 billion total parameters and 41 billion active parameters. The model also offers adjustable reasoning effort, a one-million-token context window, and a path to fine-tuning on Tinker.

2026-07-23

模型發布

OpenAI Splits GPT-5.6 Into Sol, Terra, and Luna, Standardizing a Million-Token Context Window and Agent Tool Interface

The GPT-5.6 family is now generally available, with Sol, Terra, and Luna covering maximum capability, balanced cost, and high-volume, low-cost workloads, respectively. All three models share a long-context architecture and tool interface, but their inference costs, safety behavior, and real-world quality still require separate evaluation.

2026-07-23

機器人資料與工具

Grabette Open-Sources a Low-Cost Robot Demonstration Collector That Converts Hand Manipulation into LeRobot Datasets

Pollen Robotics has released Grabette, an approximately €490 handheld device that records six-degree-of-freedom manipulation using two cameras, an IMU, and a gripper encoder. Its browser-based workflow runs SLAM, validates trajectories, and exports standard LeRobot data.

2026-07-23

程式代理評測

AgentLens Moves Coding Agent Evaluation to Production Traces, Supporting Nightly Regression Testing and Failure Diagnosis

AgentLens goes beyond checking whether an agent solves a coding task, examining the full interaction trace and actual product outcomes. The open-source benchmark can be used to compare model versions, identify behavioral regressions, and integrate evaluation into continuous integration workflows.

2026-07-23

代理評測

OmniaBench Tests General-Purpose Agents With 1,431 Cross-Scenario Tasks; Frontier Models Still Score Below 60% on Pass@1

OmniaBench introduces a cross-tool, cross-interaction benchmark for agents with explicitly defined state spaces, along with a compact 644-task hard set. The results show that planning, constraint maintenance, and post-error adaptation remain common weaknesses among frontier agents.

2026-07-23