A 10M-token context efficient agentic model.

Loading page

Model at a glance

28Bparameters

The smallest model in the comparison panel, by a wide margin.

10Mtoken context

Usable end to end, not merely addressable.

$0.15 / $1.00per 1M in / out

Below every baseline that can be bought at this length.

Benchmark overview

Every benchmark, every model, one table.

The full comparison at a glance. Each section below takes one row of this table and shows how the result was reached — where the panel diverges, and where a baseline finishes ahead.

API pricing: USD per 1M input / output tokens

| Pokee-Isaac 28B compared with five baseline models across every benchmark in the technical report, with per-token pricing for each model| Benchmark | Pokee-Isaac28B (v0)In/Out$0.15 / $1.00 | GPT-5.6 LunaAzureIn/Out$0.40 / $1.80\>272K context | Gemini 3.5 Flash LiteVertex AIIn/Out$0.30 / $2.50 | Claude Haiku 4.5BedrockIn/Out$1.00 / $5.00 | Nemotron 3 SuperAmazon Bedrock · USIn/Out$0.15 / $0.65 | Qwen 3.5 122BOpenRouterIn/Out$0.26 / $2.08 | | ---------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | --------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------ | ----------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------- | | Long context | | | | | | | | RULER256K / 512K / 1M↑ higher is better | 96.9 / 96.7 / 95.0🥇 (best in row) | 95.0 / 91.4 / 0.0\ | 94.5 / 94.6 / 29.4\ | 0.0 / 0.0 / 0.0 | 96.3 / 95.7 / 91.8s | 0.0 / 0.0 / 0.0 | | RULER2M / 4M / 10M↑ higher is better | 95.8 / 96.7 / 93.3🥇 (best in row) | 0.0 / 0.0 / 0.0 | 0.0 / 0.0 / 0.0 | 0.0 / 0.0 / 0.0 | 0.0 / 0.0 / 0.0 | 0.0 / 0.0 / 0.0 | | MRCR v2256K / 512K / 1M, 8 needles↑ higher is better | 0.607 / 0.743 / 0.500🥇 (best in row) | 0.208 / 0.173 / 0.050 | 0.474 / 0.473 / 0.205 | 0.000 / 0.000 / 0.000 | 0.145 / 0.161 / 0.067 | 0.000 / 0.000 / 0.000 | | Agentic capabilities | | | | | | | | BFCL v4overall↑ 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 2.1text-only subset↑ higher is better | 65.1% | 69.8%🥇 (best in row) | 46.5% | 34.9% | 24.4% | 46.5% | | MCP-Atlasclaim coverage↑ higher is better | 74.59% | 77.90%🥇 (best in row) | 76.67% | 56.45% | 48.95% | 70.24% | | Security | | | | | | | | DTAPattack success rate (ASR)↓ lower is better | 35.6🥇 (best in row) | 50.1 | 66.3 | 37.9 | 60.4 | 54.0 | | DTAPbenign success rate (BSR)↑ higher is better | 82.5 | 85.1🥇 (best in row) | 83.3 | 71.3 | 63.3 | 79.4 |

🥇 marks the best value in each row. ASR is attack success rate and is lower-is-safer; BSR is benign task success rate; higher is better for every other benchmark. A 0.0 / 0.000 means the model returned nothing usable at that length. \* context-overflow error at 1M. s vendor self-reported, not measured by Pokee — excluded from the row comparison. Every other figure was produced on one installation, for Isaac and each baseline alike.

Pricing: Pokee and Luna rates are as supplied/official; Gemini and Haiku use standard public API rates; Nemotron uses Amazon Bedrock on-demand pricing for US East / US West; Qwen uses the OpenRouter headline rate, which varies by provider. Haiku, Nemotron, and Qwen cannot be purchased at the context lengths above 262K that this comparison covers.

Long context

A window that is usable, not merely advertised.

RULER holds task difficulty fixed and scales only the context length, so the curve shows how far a model's usable context tracks its nominal one. Isaac is the only model in the panel that returns a score at every length.

| RULER score by context length for Pokee-Isaac 28B and five baseline models| Model | 256K | 512K | 1M | 2M | 4M | 10M | | ---------------------------------------------------------------------------------- | ------------------------------ | ------------------------------- | ------------------------------- | --------------------- | --------------------- | --------------------- | | Pokee-Isaac 28B | 96.9 | 96.7 | 95.0 | 95.8 | 96.7 | 93.3 | | GPT-5.6 Luna(Azure) | 95.0 | 91.4 | —0.0, context-overflow error | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | | Gemini 3.5 Flash Lite(Vertex AI) | 94.5 | 94.6 | 29.4\* (context-overflow error) | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | | Claude Haiku 4.5(Bedrock) | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | | Nemotron 3 Super 120B | 96.3 (self-reported by vendor) | 95.67 (self-reported by vendor) | 91.75 (self-reported by vendor) | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | | Qwen 3.5 122B | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score | —0.0, no usable score |

Pokee-IsaacBaselineVendor self-reported—No usable score (0.0)

RULER score (%, averaged across task configurations), ten samples per configuration. The 256K and 512K columns average all 13 configurations; common-words extraction is unavailable from 1M onward, so those columns average the remaining 12. \* marks a context-overflow error at that length. Nemotron's 256K–1M figures are self-reported by NVIDIA, not measured by Pokee, and are excluded from the comparison.

MRCR v2 — interference, not just depth

RULER measures how deep a model can reach; MRCR measures whether it can tell several buried targets apart. GPT-5.6 Luna scores 95.0 on RULER at 256K but 0.050 here at 1M — multi-needle disambiguation splits the panel far more sharply than single-target recall does.

Leads at every length

#### 256K

+0.133 vs. next

Pokee-Isaac0.607

Gemini 3.5 Flash Lite0.474

GPT-5.6 Luna0.208

Nemotron 3 Super0.145

Claude Haiku 4.50.000

Qwen 3.5 122B0.000

#### 512K

+0.270 vs. next

Pokee-Isaac0.743

Gemini 3.5 Flash Lite0.473

GPT-5.6 Luna0.173

Nemotron 3 Super0.161

Claude Haiku 4.50.000

Qwen 3.5 122B0.000

#### 1M

+0.295 vs. next

Pokee-Isaac0.500

Gemini 3.5 Flash Lite0.205

Nemotron 3 Super0.067

GPT-5.6 Luna0.050

Claude Haiku 4.50.000

Qwen 3.5 122B0.000

MRCR v2 (Multi-Round Co-reference Resolution), 8-needle split, 0–1 scale, all three panels on the same axis. Multiple needles are distributed through a long conversation and the model must retrieve and disambiguate a specified one, so partial recall and cross-needle interference are both penalized. Isaac leads at every measured length and the margin over the closest baseline widens as context grows. A score of 0.000 means the model returned nothing usable at that length.

Agentic capability

Four benchmarks, four different facets of agency.

Each probes something the others cannot mask: deterministic function calling, sustained multi-turn coherence, execution in a real shell, and discovery across live tool servers. Isaac leads two, places second on one, and third on one.

BFCL v4

Function calling, scored by AST and state-transition matching rather than an LLM judge.

1st of 6

Pokee-Isaac70.94

GPT-5.6 Luna70.61

Claude Haiku 4.567.52

Qwen 3.5 122B64.88

Gemini 3.5 Flash Lite64.85

Nemotron 3 Super33.13

Overall score across all 5,217 examples. The 0.33-point margin over GPT-5.6 Luna is parity, not a decisive lead — the substantive claim is matching the strongest cloud baseline while staying deployable in boundary.

τ³-bench

Multi-turn customer-service tasks against a simulated user whose requirements evolve.

1st of 6

Pokee-Isaac0.6620.820 / 3-dom

Gemini 3.5 Flash Lite0.6310.774 / 3-dom

Qwen 3.5 122B0.6110.767 / 3-dom

GPT-5.6 Luna0.5270.641 / 3-dom

Nemotron 3 Super0.4260.557 / 3-dom

Claude Haiku 4.50.4080.524 / 3-dom

Overall average across all four domains — retail, airline, telecom, and banking. The three-domain average excluding banking is shown alongside; banking is where the whole panel struggles, with the best score in that column at 0.203 and Isaac at 0.186.

Terminal-Bench 2.1

Hard, realistic tasks driven through a real shell, verified by the task's own test suite.

2nd of 6

GPT-5.6 Luna69.8%60/86

Pokee-Isaac65.1%56/86

Gemini 3.5 Flash Lite46.5%40/86

Qwen 3.5 122B46.5%40/86

Claude Haiku 4.534.9%30/86

Nemotron 3 Super24.4%21/86

Text-only subset, 86 of 89 tasks, 100-turn cap. This is the one benchmark in the report where a cloud baseline finishes ahead of Isaac — a gap of four tasks.

MCP-Atlas

Discovering and composing tools across 36 live MCP servers, with neither server nor tool named.

3rd of 6

GPT-5.6 Luna77.9%6.91 turns

Gemini 3.5 Flash Lite76.7%14.99 turns

Pokee-Isaac74.6%9.10 turns

Qwen 3.5 122B70.2%11.47 turns

Claude Haiku 4.556.5%7.40 turns

Nemotron 3 Super49.0%11.18 turns

Mean claim coverage over 500 public tasks. Isaac lands within 2.1 points of Gemini while taking roughly 60% of its trajectory length.

Security

Safest of the panel on both attack axes.

DTAP places an agent in simulated environments and measures whether injected attacks succeed. Against the five baselines run on the same installation with the judge held fixed, Isaac is the safest of six on both axes while placing third of six on capability.

| DTAP red-teaming results, macro-averaged over the 12 Linux domains, same judge for every row| Model | Direct ASRlower is safer | Indirect ASRlower is safer | Combined ASRlower is safer | BSRhigher is better | | ---------------------------------------------------------------------------------------------------- | ------------------------- | -------------------------- | -------------------------- | ------------------------- | | Pokee-Isaac 28B | 36.0best (best in column) | 35.2best (best in column) | 35.6best (best in column) | 82.5 | | GPT-5.6 Luna | 54.4 | 46.1 | 50.1 | 85.1best (best in column) | | Gemini 3.5 Flash Lite | 84.1 | 49.5 | 66.3 | 83.3 | | Claude Haiku 4.5 | 38.2 | 37.1 | 37.9 | 71.3 | | Nemotron 3 Super 120B | 80.3 | 42.0 | 60.4 | 63.3 | | Qwen 3.5 122B | 60.8 | 47.8 | 54.0 | 79.4 |

DTAP (DecodingTrust-Agent Platform) places an agent in simulated environments and measures whether injected attacks succeed. Direct ASR is the attack success rate when the harmful request is in the user prompt; Indirect ASR when it arrives through tool output or the environment. BSR is benign task success — utility. Macro-average over 12 Linux domains and 6,195 judged tasks, same judge for every row.

Isaac is the safest of the six on both attack axes, and its Direct and Indirect rates differ by 0.8 points — the tightest balance in the set. Two limits are stated plainly in the report: the indirect figures are a guards-inactive measurement, because the harness matches native tool names while DTAP's attacks arrive over MCP; and explicit refusals fired on only 1.5% of malicious tasks, so most of the current defense is incidental rather than declined. Against the broader published leaderboard of sixteen systems, Isaac places #5 on Direct ASR, #6 on Indirect, and #9 on capability — a middle-of-the-field result, and the report states it as one.

Efficiency

Prefill throughput rises with context length.

Measured under the same RULER workload whose accuracy is reported above, on a single B200-class GPU — so the serving profile is directly comparable to the capability curve rather than benchmarked on a friendlier task.

137,200peak prefill tok/s at 10M

Top of the measured sweep on a single NVIDIA B200; 42,400 at 1M.

72.9 stime to first token

For a full 10M-token prompt on that B200.

\~335decode tok/s, flat

Unchanged from 1M to 10M context.

| Serving profile under the RULER workload on a single B200-class GPU| Context | Concurrency | TTFT | Prefill (tok/s) | Decode (tok/s) | | ----------------------------------------------------------------------------- | ----------- | ------ | --------------- | -------------- | | 1M | 1 | 23.6 s | 42,400 | 335 | | 1M | 4 | 49.3 s | 81,200 | 322 | | 10M | 1 | 72.9 s | 137,200 | 337 |

Pricing

Cheaper on both meters, across a window an order of magnitude larger.

Published rates per million tokens. Three models in the panel carry a rate but cannot be bought at the context lengths this evaluation covers — no commercially available endpoint serves them beyond 262K.

| List pricing in USD per million tokens at long-context lengths| Model | Max context | Input $/M | Output $/M | | ---------------------------------------------------------------------- | ----------------------- | --------- | ------------------- | | Pokee-Isaac 28B | 10M | $0.15 | $1.00 | | GPT-5.6 LunaAzure | 1.05M | $0.40 | $1.80\>272K context | | Gemini 3.5 Flash LiteVertex AI | 1M | $0.30 | $2.50 | | Claude Haiku 4.5Bedrock | 200Knot sold above 262K | $1.00 | $5.00 | | Nemotron 3 Super 120BAmazon Bedrock · US | 262Knot sold above 262K | $0.15 | $0.65 | | Qwen 3.5 122BOpenRouter | 262Knot sold above 262K | $0.26 | $2.08 |

Add credits

Retrieved from each provider's public pricing page on 3 August 2026; long-context rates are quoted where a provider meters them separately. Nemotron uses Amazon Bedrock on-demand pricing for US East / US West; Qwen uses the OpenRouter headline rate, which varies by provider. Pokee-Isaac rates are provisional and subject to confirmation at launch. For in-boundary deployments the per-token comparison understates the difference — cost becomes a fixed function of the hardware provisioned rather than a variable function of tokens consumed.

Deployment

From a datacenter GPU down to a single consumer card.

Portability is a first-class property: Isaac is adapted to run on heterogeneous accelerator hardware, served natively through the Pokee SDK. A 28B footprint is what makes the small end of this range possible.

Datacenter1× B200-class GPU

137,200 tok/s prefill at 10M context, decode flat near 335 tok/s.

Workstation1× RTX 4090 / RTX 5090

A single consumer GPU is enough to run Isaac privately — no datacenter part required.

Client GPU1× Intel Arc Pro B70

3.6–5× the prefill and 2.3× the decode of stock llama.cpp on the same card.

Edge NPUQualcomm, Intel Panther Lake

On-device at extended context on NPU-class mobile silicon; AMD in progress.

OpenAI-compatible

Point any chat-completions client at the Pokee base URL — no new SDK to learn.

Built for tool use

A purpose-built agentic architecture for function calling and long-horizon execution.

Streaming and background

Stream over SSE, or run generation in background mode that survives a disconnect.

Deployable in boundary

Licensed to run inside a VPC, on-premises, or on-device — no request leaves your perimeter.

On-device, Intel Arc Pro B70Pokee stack vs. stock llama.cpp, same hardware

Prefill throughput

1,087–1,500 tok/s3.6–5×

Stock llama.cpp: 305 tok/s

Decode throughput

58.8 tok/s2.3×

Stock llama.cpp: 25.7 tok/s

Get started

Call the model.

Swap in your key and the base URL below. Everything else is a standard chat-completions request.

Request bodies over 16 MiB require SSE. Set `stream: true` and send `Accept: text/event-stream`. See long-context requirements.

curlPythonNode

Copy

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"}] }'

Text only

Pokee-Isaac is a text model for now. Image, audio, and video inputs are not supported.