Executive Summary
- •Maintains a 93.3% score on RULER at 10M tokens.
- •Achieves a 72.9-second time-to-first-token at 10M tokens on a single B200-class GPU.
- •Provisional pricing sits at $0.15 per million input tokens and $1.00 per million output tokens.
Community Sentiment
Key Developments & Data
Pokee AI releases Isaac 28B, a licensed foundation model featuring a 10M-token context window.
The model maintains a 93.3% score on RULER at 10M tokens, while GPT-5.6 Luna and Gemini 3.5 Flash Lite experience context-overflow at 1M tokens.
A single B200-class GPU achieves a 72.9-second time-to-first-token at 10M tokens, scaling prefill throughput to 137,200 tokens per second.
The non-open-weight release targets regulated industries with deployment inside VPCs, on-premises, and on-device via Intel and Qualcomm chips.
Provisional list pricing sits at $0.15 per million input tokens and $1.00 per million output tokens through an OpenAI-compatible API.
Zubiqo Intelligence Briefing
Get the unfiltered signal before markets open.
Top tech breakthroughs, venture funding, and market moves—synthesized into a 2-minute morning read. Zero PR fluff.
✓ 100% Free•✓ 1-click unsubscribe•✓ No spam ever
Zubiqo Strategic Assessment
Primary Impact
Regulated industries and public-sector institutions that require strictly isolated, on-premises AI deployments.
Strategic Shift
The migration of frontier-level long-context windows and agentic capabilities from cloud-only endpoints to locally hosted infrastructure.
The Ripple Effect
Enterprise software vendors will face immediate pressure to support local 10M-token models as a baseline requirement for compliance-heavy government and healthcare contracts.
This intelligence assessment is generated by Zubiqo's AI for informational purposes only.
