{"slug": "pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu", "title": "Pokee-Isaac 28B: 10M-token context, deployable on a single GPU", "summary": "Pokee-Isaac 28B, the world's first real 10M-token context frontier-class agentic model, is now deployable on a single GPU starting from RTX 4090, according to its release announcement. The model achieves 93.3% RULER at 10M tokens, up to 137K tokens/s prefill on one B200, and leads BFCL v4 and τ³-bench evaluations, with pricing at $0.15/M input and $1/M output.", "body_md": "Releasing Pokee-Isaac 28B — the world’s first real 10M-token context frontier-class agentic model, deployable on a single GPU (starting from RTX 4090 or equivalent).\nNew proprietary non-decoder-only architecture:\n• 93.3% RULER at 10M tokens\n• Up to 137K tokens/s prefill on one B200 with 10M-token context\n• Leads BFCL v4 and τ³-bench in our evaluation\n• Lowest combined attack success rate among evaluated models on DTAP security red-teaming benchmark\nPricing and deployment:\n💰 $0.15/M input · $1/M output\n🔒 Deploy in your VPC, on-premises, or on-device, with Day-0 support for", "url": "https://wpnews.pro/news/pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu", "canonical_source": "https://twitter.com/Pokee_AI/status/2084682445648216383", "published_at": "2026-08-05 16:15:53+00:00", "updated_at": "2026-08-05 16:37:55.360952+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-products", "ai-infrastructure"], "entities": ["Pokee-Isaac 28B", "RTX 4090", "B200", "BFCL v4", "τ³-bench", "DTAP"], "alternates": {"html": "https://wpnews.pro/news/pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu", "markdown": "https://wpnews.pro/news/pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu.md", "text": "https://wpnews.pro/news/pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu.txt", "jsonld": "https://wpnews.pro/news/pokee-isaac-28b-10m-token-context-deployable-on-a-single-gpu.jsonld"}}