PulseAugur
EN
LIVE 18:29:38

Pokee AI launches 28B model with 10M-token context for on-premise use

Pokee AI has released Pokee-Isaac 28B, a 28 billion parameter text-only foundation model designed for deployment within private customer boundaries. This model boasts a 10 million token context window, enabling it to maintain coherence and process extensive data without leaving a secure environment. Pokee-Isaac 28B demonstrates strong performance on agentic benchmarks, achieving 93.3% on RULER at its maximum context length and outperforming cloud-based models that falter beyond 2 million tokens. AI

IMPACT Enables secure, long-context AI applications in regulated industries and on-device scenarios.

RANK_REASON Pokee AI is a frontier lab releasing a new model with significant context window capabilities. [lever_c_demoted from frontier_release: ic=2 ai=1.0]

Read on Mastodon — fosstodon.org →

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

Pokee AI launches 28B model with 10M-token context for on-premise use

COVERAGE [2]

  1. MarkTechPost TIER_1 English(EN) · Michal Sutter ·

    Pokee AI Releases Pokee-Isaac 28B: A 10M-Token Context Agentic Model Built to Run Inside the Customer Boundary

    <p>Pokee AI released Pokee-Isaac 28B, a 28B text-only foundation model with a 10M-token context window built to run inside the customer boundary. It scores 93.3% on RULER at 10M tokens, where every baseline in its comparison panel returns 0.0 beyond 2M, and leads BFCL v4 at 70.94…

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Pokee AI has launched Pokee-Isaac 28B, a 28bn-parameter model with a 10m-token context window built to run inside customer boundaries. It scores 93.3% on RULER

    Pokee AI has launched Pokee-Isaac 28B, a 28bn-parameter model with a 10m-token context window built to run inside customer boundaries. It scores 93.3% on RULER at 10m tokens, outperforming cloud baselines that return zero beyond 2m. Pricing starts at 0.15 USD per million input to…