Pokee AI has released Pokee-Isaac 28B, a 28B parameter text-only foundation model featuring a 10M-token context window designed to operate entirely within customer-controlled boundaries. The model is available via an OpenAI-compatible API and licensed for deployment in VPCs, on-premises environments, or on-device hardware, rather than being open-weight.
- Pokee-Isaac 28B achieves 93.3% on the RULER benchmark at 10M tokens, maintaining parity with strong cloud baselines on agentic benchmarks like BFCL v4 and τ³-bench.
- It demonstrates low latency and high throughput on a single B200-class GPU, with prefill speeds reaching 137,200 tokens/s at 10M context length.
- The model records the lowest combined attack success rate (35.6) on DTAP red-teaming while maintaining 82.5% benign task success.
- It supports deployment on consumer and enterprise GPUs including RTX 4090, Intel Arc Pro B70, Core Ultra Series 3, and Qualcomm Snapdragon X2 Elite.
This release enables regulated industries such as healthcare, finance, and defense to utilize long-horizon agentic capabilities without data leaving their internal infrastructure.