全部 AI 动态

OpenRouter 上的 Guardrails 是市场上最强大的：为你的 AI 流量提供集中式安全与治理 预算限制、ZDR、模型与提供商限制、提示注入防御，以及 DLP / 敏感信息检测，分层融入你掌控的规则中！🧵

MiniMax M3 might be the most underrated coding model right now. I gave it nothing but a screenshot of a chaotic 90s GeoCities-style fan page, no HTML source, just the image + the asset files, and told it to rebuild the whole thing as a sleek Apple-style 2026 site. One shot. Through OpenCode. The result is genuinely stunning. It kept the soul (the "stevibe's HyperHome" identity, the visitor counter, the guestbook, the webmaster portrait) and translated every section into clean modern design, gradient hero, proper typography, dark theme, the works.

Blown away by MiniMax M3 I’ve been using it for free inside @orca_build with the @opencode agent Mostly UI tasks & code review so far, but so far it feels on par with Opus-4.7 Didn’t expect it to be this good. (also free while it lasts)

It’s actually crazy how few people are talking about how powerful Pixverse C1 is for text-to-VFX.We’re on a completely different level of control and cinematic quality. Here’s a prompt I pushed hard: "FPV shot: Diving through a volcano erupting beneath a thunderstorm. The camera spirals through ash clouds as lava bombs explode around it. Lightning illuminates molten rock and black smoke merging into "WRATH" before the eruption detonates outward. Elemental violence, god-tier VFX." Watch this The camera dynamics, particles, lighting… this is real pro VFX. Are you testing Pixverse C1 right now?

我现在正在一台 dgx spark 上运行 stepfun 新的 step 3.7 flash。 198b 的视觉模型，跑在一台就摆在桌上的机器上。以下是如何帮你省下大约 3 小时抓耳挠腮的加载时间，因为我已经替你抓过了。 官方 README 告诉你需要 stepfun 自己的 llama.cpp fork。你不需要，主线 ggml-org 就能跑得好好的，视觉功能什么的都行，64k。别花一个小时去构建一个 fork，结果发现不用它模型也能加载。 下面这个才是真正会吃掉你一整晚的：模型占了你约 121gb 统一内存池中的 104gb，而 spark 没有 swap。请求的上下文太大，它不会干净地崩溃，而是会静默地颠簸，内核把模型页换出、再从磁盘反复读回，循环往复，而你只能盯着“loading”发呆。 判断的信号是 read_bytes。如果它涨过 104gb 的模型大小还继续涨，同时进程内存钉在 99% 且日志毫无进展，那就是卡死了。杀掉它，它不会自己恢复。 在 q8 kv cache 上加载视觉投影器时，64k 就是 128gb 机器上的上限，超过它就会在 clip loader 里颠簸，模型加上 KV 加上视觉缓冲区全都在争抢你剩下的那约 17gb。 想要完整的 256k 上下文？去掉视觉，换成 q4 kv cache，这样就能装下，121gb 里用 114gb。这就是他们 README 没有明说的取舍：大上下文是真的，只不过得用 q4 cache，而且仅限文本。 而且它是个推理模型。问它点东西，如果你得到空白回复，不是你把它弄坏了，是你把 max_tokens 设得太低，它把整个预算都花在思考上了。答案就在 reasoning_content 里。给它留点空间。 这就是那 3 小时。确切可用的参数和权重在回复里。

Liquid's LFM2.5-8B-A1B smashed OpenAI's gpt-oss-20b on tool calling We ran both locally on a MacBook Pro M5 Max, 64GB, and gave each the same trip-planning request that only completes if the model fires all 7 tool calls - weather for 3 cities, two currency conversions, an email and a reminder Outputs: LFM2.5-8B-A1B: 4.8 GB RAM usage, 7/7 tool-calls, 266 tok/s, 6.9s OpenAI gpt-oss-20b: 11 GB RAM usage, 3/7 tool-calls, 146 tok/s, 15.0s The 8B used less than half the RAM and still fired all 7 calls, while the 20B silently dropped more than half of its own. It also ran ~2x faster, wrapping the full agentic request in 6.9s against 15s. That's what 38T training tokens buy: a 1B-active MoE that nails the agentic tool calls a model 2.5x its active size keeps dropping

Gemini 3.5 flash + Gemini managed agents api just audited a real megatron-lm ci failure inside Eigent. root cause in minutes! watch the handoff: coordinator agent plans the audit, developer agent loads the ml-failure-audit skill and gathers the evidence, then gemini agent steps in as a remote sub-agent for the heavy reasoning. gemini managed agents api and gemini 3.5 flash now live on our open-source cowork Eigent! @googleaidevs @GoogleAIStudio

🚀 The team at @Google just released the Agents API, a service for building and running custom agents inside a sandboxed Linux environment, and we built a template that gives these agents access to LlamaParse / LiteParse, enabling them to process unstructured documents automatically 📄⚡ Here’s how it works: 🔹 Configure a Git repository where data and outputs will be stored 🔹 Clone the repository into the agent sandbox 🔹 Install the LiteParse CLI, the LlamaParse SDK, and agent skills to use both 🔹 Prompt the agent with a task and watch it process documents autonomously 🤖 The result? An agent that can work directly with messy, real-world documents using LlamaParse and LiteParse within Google’s new agent runtime. Check out the GitHub repository: https://github.com/run-llama/antigravity-demo Get started with LlamaParse: https://cloud.llamaindex.ai/signup

I gave the same Go MMRPG backend spec to two coding agents and then asked Claude to judge the performance. CC (with Sonnet 4.6) 4.6 shipped 1,409 lines of clean, readable code. CyOps with @MiniMax_AI M2.7 shipped 5,271 lines of denser, layered code.

The scariest bug from a coding agent isn't the one that crashes. It's the one that runs cleanly, passes tests, and quietly produces wrong results. So we built one on @Antigravity's Gemini Managed Agents API to hunt them. Give it a repo. Get back the bugs that passed review.