NAll news

All news

Wednesday113 items

  3. Latent SpaceProduct release
    AI score85

    DeepSeek open-sources the DeepInfer inference framework, processing 155 trillion tokens per day in a single cluster

    DeepSeek has open-sourced DeepInfer, its self-developed inference framework that supports DeepSeek-V4.1-Flash, and published production data: a single cluster processes 155 trillion tokens per day, 96% input cache hit rate, 400ms first-token latency, and 230 tokens/s generation speed. The framework is customized based on vLLM and supports distributed inference, speculative decoding, SWA Bounded Replay, and structured output, with target platforms NVIDIA SM100 and Huawei Ascend 950. DeepSeek also states that components such as KV cache management and load balancing are not yet fully open-sourced, while the vLLM team says that under a matching GB200 configuration, decoding throughput is within 6% of the results published by DeepInfer.

    Same story · 29 reports

    #English text is being prepared.#Reasoning#English text is being prepared.

  4. Ayi AI NotesProduct release
    AI score70

    Open-source project LangAlpha v2.0 launches: multi-agent investment research with brokerage order placement

    The open-source project LangAlpha v2.0 has officially launched, putting financial data, research tools, multi-agent collaboration, and brokerage connectivity into a single workspace, covering markets including US stocks, Hong Kong stocks, A-shares, Japan, South Korea, Germany, and India, as well as Crypto, gold, and commodities. It offers four tiers of trading authorization, ranging from no trading at all, to manual approval for every trade, to producing a strategy plan before acting, to full autonomy, and has already integrated brokerages such as Robinhood, supporting bring-your-own API keys, Skills and MCP plugins, and self-deployment. The author notes that the open-source community has already given it more than 1,800 stars, but financial volatility and goal drift in long tasks will be amplified by automation, so it is advisable to start by requiring approval for every trade.

    Quoted postLangAlpha Chinese/CN · @ginlixai

    English text is being prepared.

    #English text is being prepared.#Agents#English text is being prepared.

  5. Latent SpaceProduct release
    AI score81

    DeepSeek open-sources the DeepInfer inference framework, processing 155 trillion tokens per day on a single cluster

    DeepSeek has open-sourced DeepInfer, the production inference framework supporting V4.1 Flash, and published production data including 155 trillion tokens processed per day on a single cluster, a 96% input cache hit rate, an average 400ms time to first token, and 230 tokens/s generation. The framework is customized based on vLLM and uses the Apache 2.0 license, supports distributed inference, speculative decoding, SWA Bounded Replay and structured output, and targets NVIDIA SM100 and Huawei Ascend 950. DeepSeek says components such as KV cache management and load balancing are not fully open-sourced because they are coupled with internal infrastructure, while the vLLM team reports that under matched GB200 configurations, decoding throughput differs from DeepInfer's published results by less than 6%.

    Same story · 29 reports

    #English text is being prepared.#Reasoning#English text is being prepared.

  13. Shao MengProduct release
    AI score67

    Databricks Releases MemCurrent Long-Term Memory Architecture, Internal Benchmark Pass Rate Tripled

    Databricks has released MemCurrent, a persistent memory architecture for production Agents, combining episodic memory with derived memory and adopting three integration approaches: Plugin, Tool, and offline dreaming. Based on 114 internal test cases selected from more than 12,000 real Genie sessions, the pass rate rose from 22.3% without memory to 68.4%, while also being 1.35 times faster and 1.44 times cheaper. On the public benchmarks LoCoMo, LongMemEval, and ConvoMem, the average score was 88.1%, comparable to 87.8% with full context, but with about 12 times fewer input tokens per question (8.4K vs. 103K). The architecture has already been used in Agent Bricks Memory and Genie.

    Quoted postDatabricks · @databricks

    Introducing MemCurrent, a persistent memory architecture designed to help production agents stop starting from zero. MemCurrent combines episodic memory with derived memory across three access patterns: • Plugins to log and load context automatically • Tools the agent calls on demand • An offline “dreaming” pipeline that consolidates memory in the background Agents carry forward user preferences, prior corrections, metric definitions and successful workflows instead of rediscovering them every session. On our internal benchmark built from real enterprise data, MemCurrent triples the agent's pass rate while using 31% fewer tokens and running 26% faster. It's built into Agent Bricks' Memory offering and already powers memory in Genie. Read the research: https://t.co/AYGXOqrhkO

    #Product updates#Agents#RAG

  14. TiboProduct release
    AI score52

    OpenAI API supports 12-hour prompt caching; caches are retained long-term for logged-in Codex and Work users

    The OpenAI developer account announced that the API can now opt into 12-hour prompt caching, so shared instructions, tools, and context can stay cached longer, and even when requests are hours apart they are billed at the lower cached-input rate. Tibo added that for logged-in Codex and Work users, caches are retained indefinitely, so returning to an old conversation or continuing hours later incurs no extra cost and cache hits are guaranteed.

    Quoted postOpenAI Developers · @OpenAIDevs

    You can now choose 12-hour prompt caching in the API. Persistent agents, like dots, reuse context throughout the day. You can keep shared instructions, tools, and context cached longer and pay the lower cached-input rate, even when requests are hours apart.

    #Product updates#English text is being prepared.#OpenAI

  16. testingcatalogProduct release
    AI score52

    English translation is being prepared.

    Quoted postAnusha Ramesh · @anurahk

    I launched the first image and video generation models at Google, and kept asking why all AI designs look the same. AI has brought us the WORST era of design. It made content creation effortless, but it also made 90% of it look painfully generic. My co-founder and I spent 2 years building in stealth to change that. Meet Vyb, the first AI design model that understands your taste like a designer would. RT + comment your website URL and we’ll make your first design.

    #Model releases#Image generation#Agents

  17. testingcatalogProduct release
    AI score60

    English translation is being prepared.

    Quoted postEagle · @EagleCorp

    EAGLE-4 is here. From EAGLE-1 to EAGLE-3, we’ve kept making inference faster. EAGLE-4 is our fastest yet. 961.6 tokens/s on @Alibaba_Qwen’s Qwen3.8-27B with TokenSpeed (BF16, 4× @NVIDIA B300, TP4, one request, T=0). Github: https://t.co/GwocZk1UTU https://t.co/GawgFNXKWj

    #Model releases#Open-source ecosystem#Reasoning

  19. UCSD Hao AI LabFirst-partyProduct release
    AI score55

    Hao AI Lab's FastH3 OmniRef Now Supported by ComfyUI

    ComfyUI now supports Hao AI Lab's FastH3 OmniRef, a distilled version of Minimax H3 that runs on local hardware, supports multimodal reference video generation, and strikes a balance between quality and speed. ComfyUI states that open weights and a cloud entry point are available via the links it provides.

    Quoted postComfyUI · @ComfyUI

    FastH3 OmniRef by @haoailab's is now supported in ComfyUI. This distilled version of Minimax H3 allows multi-modal reference generation of videos. It runs locally on your hardware and provide great balance between quality and speed. Click the links below for open weights and cloud ⬇️

    #Product updates#Video#Open-source ecosystem

  20. ChatGPTFirst-partyProduct release
    AI score60

    ChatGPT launches Locked Chats feature

    ChatGPT introduces Locked Chats, which moves conversations out of regular history and search and into a protected space. On web and desktop, they are unlocked with a PIN; on supported mobile devices, a PIN, Face ID, Touch ID, or device unlock can be used. New messages in locked conversations do not update the memory used by regular conversations. The feature is rolling out starting today to the Free, Go, Plus, Pro, Business, Enterprise, and Edu plans.

    Same story · 4 reports

    #Product updates#OpenAI