PulseAugur / Brief
EN
LIVE 19:58:13

Brief

last 24h
[6/6] 224 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. RESEARCH · Mastodon — sigmoid.social English(EN) · · [2 sources]

    Subquadratic – Introducing SubQ 1.1 Small https:// subq.ai/subq-1-1-small-technic al-report # HackerNews # Subquadratic # SubQ # Small # Update # TechNews # AI

    Subquadratic Inc. has released SubQ 1.1 Small, an updated version of their language model. The release includes a technical report detailing the advancements in this new iteration. Further specifics on the model's capabilities or improvements are not detailed in the provided information. AI

    IMPACT Subquadratic Inc. has released an updated version of its SubQ language model, SubQ 1.1 Small, with accompanying technical details.

  2. SIGNIFICANT · TLDR AI English(EN) ·

    GPT-5.5 Instant ⚡, SubQ 12M context 🧠, Gemini Flash upgrades 🚀

    OpenAI has launched GPT-5.5 Instant, an update to its default ChatGPT model, focusing on enhanced factual accuracy and reduced hallucinations. Concurrently, Subquadratic has introduced a new AI model boasting a 12-million-token context window, reportedly outperforming GPT-5.5 on retrieval benchmarks and aiming for a 50-million-token window soon. Meta is also developing a personalized AI assistant powered by its Muse Spark model, designed to perform everyday tasks and learn with less human intervention, with a target launch before the end of the year. AI

    GPT-5.5 Instant ⚡, SubQ 12M context 🧠, Gemini Flash upgrades 🚀

    IMPACT New models with significantly larger context windows and improved accuracy will likely accelerate adoption of complex AI applications and agentic systems.

  3. SIGNIFICANT · Mastodon — mastodon.social English(EN) ·

    The context window has been shattered: Subquadratic debuts a 12M token window https://thenewstack.io/subquadratic-12-million-context-window/ # HackerNews # Tech

    Subquadratic has unveiled a new model with a 12 million token context window, significantly expanding the amount of information an AI can process at once. This breakthrough shatters previous records for context window size. The development promises to enable more complex and nuanced AI applications by allowing models to retain and analyze much larger amounts of data. AI

    IMPACT Enables AI to process and analyze vastly larger datasets, potentially unlocking new capabilities in complex reasoning and long-form content generation.

  4. RESEARCH · Mastodon — fosstodon.org English(EN) ·

    A Miami-based startup called Subquadratic came out of stealth last week with a single claim that’s either the most important architectural shift since the 2017

    Subquadratic, a Miami-based startup, has emerged from stealth claiming to have developed the first Large Language Model (LLM) that does not utilize quadratic attention. This architectural innovation reportedly enables the model to process a context window of 12 million tokens at a significantly reduced cost compared to existing frontier models. AI

    A Miami-based startup called Subquadratic came out of stealth last week with a single claim that’s either the most important architectural shift since the 2017

    IMPACT Potential to drastically lower inference costs for LLMs with extremely long context windows.

  5. RESEARCH · dev.to — Anthropic tag English(EN) ·

    1,000x Claim, No Independent Proof: Subquadratic Architecture

    Subquadratic, a new AI startup, has emerged from stealth claiming its novel subquadratic architecture can reduce attention compute by nearly 1,000x for very large context lengths. The company launched its first model, SubQ 1M-Preview, and three private beta products, including an API and a coding agent, built on this architecture. However, at launch, Subquadratic had not published independent research to validate its significant claims, leading to a mix of curiosity and demands for proof from the AI community. AI

    IMPACT Potentially disruptive if claims are validated, offering significant cost reductions for long-context AI applications.

  6. TOOL · dev.to — LLM tag English(EN) ·

    12 million tokens, linear cost: Subquadratic's bet against the attention tax

    Subquadratic, a startup with 11 PhD researchers, has launched a new model featuring its Subquadratic Selective Attention (SSA) architecture, which claims to scale linearly with context length. This innovation allows for a 12-million-token context window, aiming to overcome the quadratic cost limitations of traditional dense attention mechanisms in LLMs. Early benchmarks show competitive performance against models like GPT-5.5 and Claude Opus on tasks such as MRCR v2 and SWE-Bench, with significantly faster inference speeds. AI

    12 million tokens, linear cost: Subquadratic's bet against the attention tax

    IMPACT Linear scaling in compute and memory with context length could significantly reduce the cost and improve the ROI of RAG and agentic decomposition.