Pokee-Isaac
GPT-5.6 Luna
Gemini 3.5 Flash Lite
Claude Haiku 4.5
Nemotron 3 Super
Qwen 3.5 122B10M tokens with real recall
Hold codebases, research corpora, histories, and multi-step work in one active context—without losing accuracy deep into the window.
Loading page
Pokee Isaac Model Profile10M
maximum context
28B
parameters
93.3
RULER at 10M
1×
RTX 4090
Throughput figures are peak results on 1 NVIDIA B100.
Deploy across the silicon ecosystem
Every figure produced by Pokee AI on one installation, for Isaac and for each baseline alike, against the cost-optimized cloud tiers and the leading open-weight models.
| Benchmark | Pokee-Isaac | GPT-5.6 Luna | Gemini 3.5 Flash Lite | Claude Haiku 4.5 | Nemotron 3 Super | Qwen 3.5 122B |
|---|---|---|---|---|---|---|
| RULERat 10M tokens higher is better | 93.3 (best in row) | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 |
| MRCR v2at 1M, 8 needles higher is better | 0.500 (best in row) | 0.050 | 0.205 | 0.000 | 0.067 | 0.000 |
| BFCL v4function calling higher is better | 70.94 (best in row) | 70.61 | 64.85 | 67.52 | 33.13 | 64.88 |
| τ³-bench4-domain average higher is better | 0.662 (best in row) | 0.527 | 0.631 | 0.408 | 0.426 | 0.611 |
| Terminal-Bench 2.1shell execution higher is better | 65.1% | 69.8% (best in row) | 46.5% | 34.9% | 24.4% | 46.5% |
| MCP-Atlastool orchestration higher is better | 74.59% | 77.90% (best in row) | 76.67% | 56.45% | 48.95% | 70.24% |
| DTAPattack success rate lower is better | 35.6 (best in row) | 50.1 | 66.3 | 37.9 | 60.4 | 54.0 |
Trusted across the enterprise
Keep capability, cost control, and data sovereignty in the same system.
Hold codebases, research corpora, histories, and multi-step work in one active context—without losing accuracy deep into the window.
Built for function calling, multi-turn tool use, and long-horizon execution rather than single-turn chat.
Run in your VPC, on-prem, or on-device—from cloud GPU to edge NPU.
Bring the model to enterprise data and regulated workflows instead of sending sensitive context to a third-party cloud.
Use Pokee-Isaac as the reasoning engine behind agents that must remember, decide, and act over long horizons.
Plan changes across large codebases, call tools, test results, and retain project context across long runs.
Read large corpora, cross-check evidence, and build source-grounded reports without fragmenting the workflow.
Keep legal, finance, procurement, and support data inside enterprise-controlled infrastructure.
Run high-volume, repeated agent tasks with pricing designed for production rather than one-off demos.
Choose pay-as-you-go API access for individual developers, or contact Pokee for enterprise deployment and support.
Use the OpenAI-compatible endpoint to add Pokee-Isaac to an existing application or start a new agent workflow.
pokee-isaac$0.15 / 1M input tokens
$1.00 / 1M output tokens
No long-term commitment.
Private deployment, volume pricing, security review, integration support, and custom infrastructure options.
Contact PokeeRequest bodies over 16 MiB require SSE. Set stream: true and send Accept: text/event-stream. See long-context requirements.
curl https://api.pokee.ai/v1/chat/completions \ -H "Authorization: Bearer pk-..." \ -H "Content-Type: application/json" \ -d '{ "model": "pokee-isaac", "messages": [{"role": "user", "content": "hello"}] }'