搜索“RAG”

cite the cells they came from, and charts can drop directly into the sheet beside the data. Write formulas and run scenarios Describe the outcome and Grok handles the syntax — averages, sorts, and fills land as normal edits in the formula bar. Test a downside case, rerun a forecast, or close out month-end without rebuilding the sheet by hand. The add-in can also leverage connectors ; ask it to pull context from recent emails or files in SharePoint or Google Drive. Grok for Excel is a free Micro

匹配字段正文

ve no data stores in the critical path of sandbox creation at all, which improves scalability and reliability. While we do need to write sandbox metadata and results to durable storage, we do so largely asynchronously; Other than sandbox creations, we have no RPCs which are O(sandboxes). Workers batch control messages for multiple sandboxes together in single RPCs, in the spirit of ideas from data-oriented design The result is that the sandbox creation path requires only two network hops and one

匹配字段正文

ited to partner with the Thinking Machines team to bring Day-0 support for Inkling to SGLang and Miles, with dedicated optimizations for its new architecture and broad feature coverage, and with Modal to train a DFlash draft model for Inkling. Highlights Inkling is a 975B-parameter multimodal model with a context window of up to 1M tokens, featuring short convolution, attention with relative positional embedding, and a novel MoE design with a shared-expert sink; SGLang delivers up to 71.7k tok/s

匹配字段正文

loyments that only a custom port can target to extract maximum inference performance. Parallelization across GPUs, compilation, fused kernels, and many more, all contribute to leveraging your hardware to achieve ultra-fast inference. When model authors wanted the absolute best performance, they were still writing custom vLLM implementations. The latest iteration of the transformers modeling backend for vLLM dynamically applies inference specific layer fusions at runtime to match the speed of cus

匹配字段正文

ching, no manual environment setup, and no permission troubleshooting. More from this author Record, train, and deploy from one place with Strands Agents, LeRobot, and Hugging Face Storage Buckets 34 August 13, 2026 From the Hugging Face Hub to robot hardware with Strands Agents and LeRobot 17 June 17, 2026 Community Edit Preview Upload images, audio, and videos by dragging in the text input, pasting, or clicking here . Tap or paste here to upload images Comment · Sign up or log in to comment +9

匹配字段正文

d Apr 10 • 457 • 8.29k • 5 More Articles from our Blog Grabette: an open system to record robot-manipulation data +2 64 July 21, 2026 lerobot robotics LeRobot v0.5.0: Scaling Every Dimension +6 45 March 9, 2026 Community Edit Preview Upload images, audio, and videos by dragging in the text input, pasting, or clicking here . Tap or paste here to upload images Comment · Sign up or log in to comment +73 Models mentioned in this article 3 Datasets mentioned in this article 1 lerobot/robotwin unified

匹配字段正文

ut this: kernels CLI kernel-builder CLI This improved CLI experience also lets us cater to the rise of agentic kernel development in a better manner. More on this later . More coverage of frameworks and backends We have extended support for frameworks, the most visible changes are: We added support for the Torch Stable ABI to kernels and kernel-builder. The Torch Stable ABI allows kernel developers to target a particular Torch version or any version that is released after it for roughly two year

匹配字段正文

hey can't act on. Some teams are seeing 10–100x more security findings—and many are false positives. Security teams are stuck with tools that haven't caught up. Scanners offer coverage at scale but miss critical exploits because they can't reason about business logic and discover chained vulnerability attacks. AI security tools often struggle to reason effectively about the whole codebase. None of them can validate which findings are actually exploitable or write a fix. Devin Security Swarm brin

匹配字段正文

o CI stays green. To make sure the missing sender id was a real failure and not a broken test, it added a control: on the same /chat-stream request, an unrelated field saved to storage correctly while the sender id did not. The storage path worked; only the change's field was dropped, pointing the failure at the PR rather than the pipeline. Earning trust, and retaining it We made a handful of deliberate decisions about what goes into a report, each in the interest of earning the trust of the aut

匹配字段正文

ioned in this article 1 More Articles from our Blog +4 35 August 6, 2026 guide partnerships open-source-collab Run AI workloads on any cloud, store on Hugging Face: zero-egress storage with SkyPilot +1 35 July 7, 2026 Community We even crossed 10k Reachy Minis in the wild 😄 \n","updatedAt":"2026-07-01T19:48:14.006Z","author":{" id":"67f8d29d5ae1f08f99dc0811","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/AkX9RbZqzzQlz797iJeqL.jpeg","fullname":"Remi Fabre","name":"Re

匹配字段正文

ng on CPU, then hand off to ESMFold2 or a similar model for structure prediction on GPU. With Modal, you request a GPU for that one function, not for the whole pipeline. Shared storage across jobs. Moving around things like sequencing datasets, protein databases, and model checkpoints between pipeline steps is expensive. With Modal Volumes, it's done once and available across every job, whether you're running one query or thousands. Reproducible environments. Modal Images make it easy to define

匹配字段正文

t coding benchmark that measures both code correctness and quality. Devin Fusion: Frontier Performance at up to 60%† Lower Cost Score on FrontierCode 1.1 Extended Benchmark and average cost per task Data last updated 8/7/2026 with current frontier models † We initially reported a 35% cost reduction at publication. On the latest FrontierCode 1.1 Extended data (updated 8/7/2026), Fusion is up to 60% cheaper, as shown in the updated charts. Largest efficiency gains are seen in implementation-focuse

匹配字段正文

June 29, 2026 Gemma 4 is now significantly faster in Ollama 0.31. On Apple Silicon, it generates tokens nearly 90% faster on average across a coding-agent benchmark. The speedup is on by default, and it does not change the model’s output: The speedup comes from multi-token prediction (MTP). Gemma 4 ships with a small, fast draft model that runs alongside the main model and proposes the next several tokens. The main model then verifies that proposal in a single pass and keeps the tokens it agrees

匹配字段正文

Dev modes, Motion is the newest layer of a connected experience on the canvas. Switch any frame to Motion mode, and the timeline appears alongside your design. Timeline control: Drag layers to adjust timing; scrub to preview any moment; and keyframe position, scale, rotation, and opacity independently. Enable auto keyframing to record every change you make while the playhead is moving. And with time-based comments on the canvas, anyone can point to a specific moment in the animation, bringing t

匹配字段正文

ning for computer use in Gemini 3.5 Flash. We’re also releasing two optional enterprise safeguard systems that enable enterprises to: Taking a “defense-in-depth” approach, we encourage developers to combine these features with secure sandboxing, human-in-the-loop verification and strict access controls. Additional information on safety measures can be found in our best practices documentation. We are already seeing customers drive value with computer use. Here’s what some of them have to say: To

匹配字段正文

plexity, code smells, potential security vulnerabilities — are way less prevalent. And the code review bot makes sure new functionality doesn’t get added without tests, so our coverage has increased by nearly 25%. The only metric we’ve found to decrease is the Maintainability Index. But lines of code count against maintainability, regardless of their quality, and over the same interval the repository grew 144% from 61 KLOC to 148 KLOC. Closing the loop: the iterate skill Setting up the verificat

匹配字段正文

algorithm development from data semantics. There are a number of benefits here. For instance, large terabyte-scale model weights or datasets don’t need to be moved around from storage to development servers. They don’t even need to be loaded from disk or pass through CPU RAM — they can be generated directly on device! Mocking speculation is trickier. The speculator is its own ML model, so you can’t just run it on random tensors, or the acceptance lengths will plummet. The speculator is operatin

匹配字段正文

ination rate among frontier models, offers 1-million-token context window, and supports configurable reasoning efforts (none, low, medium, high). Now, Amazon Bedrock users can leverage the model through Bedrock’s secure and reliable inference engine. Leading Performance in Enterprise Grok 4.3 is one of the strongest models available today for building enterprise-ready AI agents: 1 on Artificial Analysis Omniscience benchmark , achieving the lowest hallucination rate among frontier models. 1 on A

匹配字段正文

rmed 6.5 percentage points worse than GPT-5 Codex on the official leaderboard. To diagnose the regression, we used Docent to filter for tasks where GPT-5.1 Codex receives lower average reward across rollouts than GPT-5 Codex. On these tasks, Docent selected one failed run from GPT-5.1 Codex, paired it with the successful runs from GPT-5 Codex, and identified where the models’ strategies diverged. Read the analysis plan. Conclusion Analysis plans are now available to all Docent users. To try them

匹配字段正文

n each TDN Pod. By the way, don’t underestimate that less than 1 microsecond for a single hop over TDNLink. Anand says that a hyperscaler told them that with NVLink Switch, the average latency in the real world with real AI workloads is 10 microseconds to 13 microseconds. NVSwitch is very wide, but it ain’t necessarily as fast as we are all probably assuming. (We’d love to confirm this number with Nvidia, of course.) What we would call a chassis, Tensordyne calls a TDN Pod, and it looks like thi

匹配字段正文