Pokee Launches 28B Agentic Model with 10M Token Context, Runs on a Single RTX 4090
Scobleizer · x · 2026-08-05
Startup Pokee AI launched the Pokee-Isaac 28B model, aiming to combine ultra-long context with agentic capabilities at a very small parameter scale.
Key Features:
- Ultra-Long Context: Supports a usable context of 10 million tokens (scoring 93.3 on the RULER benchmark).
- Low Deployment Barrier: With only 28B parameters, it runs on a single RTX 4090 with significantly optimized VRAM usage for KV cache.
- Agentic Performance: Excels in function calling (BFCL) and tool orchestration, while showing strong security (low DTAP attack success rate).
- High Cost-Effectiveness: Priced at $0.15/M input and $1.00/M output tokens. Deployable in VPC, on-prem, or on edge devices.
Related event: Pokee-Isaac 28B Launches: 10M Context on Single GPU, Tops BFCL v4(5 posts)→
More from coding & agent
- Goodfire Launches Silico: AI Agents Can Autonomously Run Experiments for Days — mathildepapillo · 2026-08-05
- You.com Web Search API Adds x402 Support for Autonomous Agent Payments — kleffew94 · 2026-08-05
- Retrospective: Early LLM Agentic Security Case Shows AI Immediately Scanning Network with Shell Access — moyix · 2026-08-05
- Hands-on with Latest Coding Models: GPT-5.6 Regresses, Grok-4.5 Wins — sergeykarayev · 2026-08-05
- Cloudflare kicks off Agents Week with full-stack agent updates and Connect conference discount — craigsdennis · 2026-08-05
- Goodfire Launches Silico Platform for Frontier AI Research — leland_mcinnes · 2026-08-05