Today on the blog, we discuss a pathway for the second life of phones through the exploration of “phone cluster computing”, which can directly reduce the environmental footprint of computing by avoiding the need for further raw material extraction. More →goo.gle/4aJe5vO

Introducing 𝗢𝗺𝗻𝗶𝗴𝗲𝗻𝘁, a meta-harness to combine, control, and share your agents. The best teams already mix models and harnesses and design loops that drive teams of agents. No single harness can keep up with that alone. So we built the layer above — we call it a 𝗺𝗲𝘁𝗮-𝗵𝗮𝗿𝗻𝗲𝘀𝘀. Omnigent sits above the tools you already use, Claude Code, Codex, Pi, and your own agents, and gives them one shared layer: • 𝗖𝗼𝗺𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻: combine models, harnesses, and techniques without rewriting code, and switch between them with one-line changes • 𝗖𝗼𝗻𝘁𝗿𝗼𝗹: stateful, data-centric policies and cost budgets enforced at the meta-harness layer, not via prompts — let agents run without watching them • 𝗖𝗼𝗹𝗹𝗮𝗯𝗼𝗿𝗮𝘁𝗶𝗼𝗻: share a live agent session via URL with full history, so teammates can review, comment, and steer in real time Every session is reachable from a terminal, the web, a desktop app, and your phone. We built Omnigent for our own use at Databricks and are open sourcing it under Apache 2.0. databricks.com/blog/introduc…

Introducing the Fusion API, the smartest compound model in the market. Fusion achieves Fable-level intelligence at half the price. How it works 👇

SITUATION DETECTED: The city of Rio de Janerio has post-trained a model. Based on Qwen 7/2, Rio 3.5 Open 397B adds SwiReasoning on top of the base Qwen model — a framework that dynamically switches between standard chain-of-thought and latent-space reasoning, guided by entropy-based confidence signals, so the model only "thinks out loud" when it needs to and otherwise reasons silently in hidden space for better token efficiency.

NEW: Inside the 24-hrs before WH slapped export controls on Anthropic - Last Thursday, Amazon CEO Andy Jassy raised concerns about Fable jailbreak to Trump admin - Friday AM, Sean Cairncross, Bessent, Susie etc. held WH call to discuss - Then White House started reaching out to Anthropic to speak with Dario Amodei, who was at a wellness retreat. - When Amodei was finally available past 1pm, he had three tense phone calls with a combo of ppl including Cairncross, Bessent, Lutnick, Kessler, Will Scharf, Richard Walters, and Walker Barrett. -Amodei tried to clear up what he assumed was a misunderstanding. He defended the guardrails and distinguished between universal and non-universal jailbreak - Cairncross and Bessent were unmoved and asked Amodei to take down Fable and work with the admin to fix the vulnerabilities. (A WH official said Amazon’s findings were run past the NSA and they felt they had “proof.”) - Amodei asked for more time and info, but he made no commitments to pull the model - Bessent told Amodei directly at one point that he was making a “bad decision” - By Friday evening, the Trump admin imposed its export controls. - “Export controls were a last resort after begging them for hours to work with us,” senior WH official said. W/ @cheyennehaslett politico.com/news/2026/06/13…

Game theory from here is super interesting: Original Mags (Google, Amazon, Microsoft, Meta) now have a serious non-zero opportunity to tank the frontier labs. Go to the government, kneecap the labs’ motion of putting the latest models out in the wild, become the trusted gatekeeper between the labs and the public at large (including internationally) by having the labs go through their clouds (AWS, GCP, Azure) and implement strict KYC to seal the deal. The frontier labs should have seen this coming years ago and implemented a robust KYC for just this moment. The fact they didn’t is kind of concerning. Why did they not do it? Best guess is because it would have changed the run-rate revenues (downward) which would have then changed funding dynamics - lower valuations, more dilution, less secondary. A valuation reset may happen now anyways, except the labs may end up with less control and more restrictions at the end of it. At the same time, everyone is already clamoring about token prices of the old models from the labs anyways… This couldn’t be a better setup for open source and neoclouds. Big question is can they meet the moment? There are too few of them and their progress seems sporadic at best.

New server tool: Subagent 🤖 Your model can now delegate focused sub-tasks to a smaller, cheaper, faster model mid-generation. The big model orchestrates, the subagent executes. The subagent can use any model on OpenRouter.

Today we're releasing the first results for AA-AgentPerf, our new agentic inference benchmark: initially covering DeepSeek V4 Pro across NVIDIA Blackwell, Hopper, and AMD. AA-AgentPerf is the first benchmark built for agentic inference. We use real, long-context agentic coding trajectory data as the workload, and inference with real production optimizations such as KV cache reuse and speculative decoding, leading to the most realistic evaluation of inference performance available today. AA-AgentPerf’s lead metric is Agents per Megawatt. In a power-constrained world, this answers the most relevant question for AI infrastructure providers - “how many real agents can I deploy per unit of power available?”. First results for DeepSeek V4 Pro (at the easiest defined service level of 20 tokens/s and 10s TTFT): ➤ GB300 (rack-scale, disaggregated): 61,354 Agents/MW ➤ B300 (single node, disaggregated): 21,053 Agents/MW ➤ MI355X: 3,551 Agents/MW ➤ H200: 2,594 Agents/MW Further AA-AgentPerf details: ➤ Real agent workloads, beyond synthetic queries: AA-AgentPerf replays real coding agent trajectories where our agents used up to 200 turns and worked with sequence lengths >100K tokens - the workloads that matter in 2026 ➤ Production optimizations allowed: KV cache reuse, speculative decoding, and prefill/decode disaggregation are all permitted, with accuracy verification to control for quality loss - we want results to reflect what real deployments actually look like ➤ Lead metric is Agents per Megawatt: simultaneous agents supported at production performance targets (e.g. 20 tokens/s per user, ≤10s TTFT) per megawatt consumed. Agents per TCO and $/hr will be supported soon Key findings: ➤ Rack-scale disaggregated inference (GB300) is ~3× more power-efficient than single-node Blackwell (B300), and similarly ahead in raw agents per GPU ➤ Blackwell represents a large generational step over Hopper in both power efficiency and raw compute per GPU ➤ In this test, NVIDIA's Blackwell systems currently lead AMD MI355X by a clear margin. Important context: our MI355X configs are approximately two weeks older than our Blackwell configs and couldn’t stably use speculative decoding. MI355X power draw under heavy load is also well below TDP, indicating there is much room to improve on DeepSeek V4 Pro, which we will measure and publish in the coming weeks ➤ Config and inference framework version matter enormously - we've seen meaningful improvements daily since the DeepSeek V4 Pro release and look forward to tracking performance over time AA-AgentPerf is a live benchmark and we publish results on a rolling basis as submissions come in. Some of the new features coming in v1.1: more models (gpt-oss-120b), more hardware (GB200, B200, H100, MI300X), better AMD configurations, $/hr and cost-per-task normalization, Agents per TCO, and performance tracking over time.

Intelligence should be open, accessible, and ready to build with, empowering every developer, everywhere. GLM-5.2 is now available to all GLM Coding Plan users, including Lite, Pro, Max, and Team plans. docs.z.ai/devpack/latest-mod… As our new flagship model, GLM-5.2 delivers powerful coding capabilities, usable 1M-context support, and continued strengths in long-horizon tasks. API and Chatbot services will launch next week. The model will also be officially open-sourced next week under the MIT License. The future of AI is open, and it belongs to the people.

Sandboxes are all the rage (Modal, E2B, AWS, ..). Most AI teams pay a >4x markup to run sandboxes on someone else's machines. Introducing SkyPilot Sandboxes — Run BYOC sandboxes on your own clusters. • 50,000 sandboxes on a single cluster • Sub-second launches with warm pools • Great for RL rollout (keep sandbox clusters close to GPUs) Benchmark shows @skypilot_org Sandboxes are 4-10x cheaper than Modal at lower latency. Full results in blog.

After encouraging staff to prove their “AI-driven impact,” Meta is now moving to cap employee token usage and steer workers toward in-house tools. Full story: thein.fo/4vbt2PE

OpenAI was subpoenaed by a coalition of state attorneys general for documents covering a wide range of its activities and impact on users on.wsj.com/4utkyCd

Executives and employees alike are struggling with Meta's chaotic AI strategy, according to sources and internal discussions reviewed by WIRED. wired.com/story/mark-zuckerb…

Today we're releasing ZONOS2, our next-generation real-time TTS model with high-fidelity voice cloning. ZONOS2 is the most expressive open-source TTS model, released under Apache 2.0 and available on Zyphra Cloud on @AMD. 🧵

🚀 Introducing Gemini-SQL2, our breakthrough text-to-SQL capability powered by Gemini 3.1 Pro! We've achieved state-of-the-art results on the highly competitive BIRD benchmark, translating natural language into execution-ready SQL queries. 🧵👇

金融科技独角兽 Ramp 发布了针对前沿 AI 编码智能体的私有测试基准 Ramp SWE-Bench。Ramp SWE-Bench 包含 80 个源自 Ramp 真实生产环境的后端开发任务，旨在解决公共评估数据集因模型预训练而导致的数据泄露与指标饱和问题。 测试任务均提取自 Ramp 内部 AI 编码助手 Inspect 成功部署到生产环境的真实拉取请求（PR）。每个任务重构了 PR 提交前的基础代码库，保留合并的代码与测试作为黄金标准，并提取工程师在 Inspect 对话中的原始意图作为提示词。评估采用 mini-swe-agent 沙箱环境，以单次运行成功通过所有测试且不破坏现有功能（pass@1）作为通过标准。 根据公布的 14 款模型横向评测结果，Anthropic 最新推出的 Claude Fable 5 以 87.5% 的解决率高居榜首。Claude Opus 4.7 和 GPT-5.5 并列第二，解决率均为 83.75%。Claude Opus 4.8 虽然解决率为 77.5%，但单次任务平均运行时间缩短至 8 分 30 秒，单次运行成本仅为 1.09 美元，不足 Fable 5 运行成本的 40%，展现出极佳的成本效益。 测试数据还揭示了不同模型在价格与性能间的权衡。国产模型 Kimi K2.6 与 GLM 5.1 解决率相近，分别为 72.5% 与 71.25%，但 Kimi K2.6 的平均成本为 0.69 美元，比 GLM 5.1 便宜约 34%。在轻量级模型中，GPT-5.4 Mini 以 58.75% 的解决率领先 Claude Haiku 4.5 约 10 个百分点，且平均成本和步骤数均减半。随着模型走向应用，性能、速度与成本的权衡正成为工程落地时的关键考量。

Today we’re releasing Ramp SWE-Bench: a private, production-grounded coding benchmark created from real engineering problems we've faced at Ramp.

French startup Mistral AI is in talks to raise around €3 billion ($3.5 billion) at a valuation of roughly €20 billion, sources say bloomberg.com/news/articles/…