PulseAugur
EN
LIVE 19:01:51

Together AI launches Kimi K3 with 1M context window for agent workflows

Together AI has launched Kimi K3, an open frontier model designed for complex agentic workflows. This model boasts 2.8 trillion parameters and a 1 million token context window, with capabilities including native vision for coding assistance, repository navigation, and terminal tool use. Together AI is offering a production path for these long-running, tool-heavy workloads on their US-based infrastructure with a zero data retention policy. AI

IMPACT Enables complex, long-running agentic workflows with a large context window, potentially accelerating development in areas like coding and research.

RANK_REASON The cluster announces a new open frontier model (Kimi K3) with significant technical specifications (2.8T parameters, 1M context window) and its availability on a platform (Together AI), aligning with the criteria for a frontier release. [lever_c_demoted from frontier_release: ic=2 ai=1.0]

Read on X — Together (inference / OSS) →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

Together AI launches Kimi K3 with 1M context window for agent workflows

COVERAGE [2]

  1. X — Together (inference / OSS) TIER_1 English(EN) · togethercompute ·

    Together gives developers an efficient, high-throughput production path for K3’s long, tool-heavy agent workloads, hosted on Together AI’s US-based infrastruct

    Together gives developers an efficient, high-throughput production path for K3’s long, tool-heavy agent workloads, hosted on Together AI’s US-based infrastructure with zero data retention. Start building: https://t.co/ueeSGi73te

  2. X — Together (inference / OSS) TIER_1 English(EN) · togethercompute ·

    K3 on Together AI is built for long-running agent workflows:

    K3 on Together AI is built for long-running agent workflows: → 2.8T parameters and a 1M context window → Native vision for screenshot-guided coding → Repository navigation and terminal tool use → Complex research and knowledge workflows