[ Model ]

NextGen. MMLU 90.6%

Our most advanced model — designed for the future.

Deep reasoning
Multimodal
Large context
Try in QuarkAPI docsPricing
Context
1M

Window

Latency
~20ms

Typical TTFT

Highlight
90.6%

MMLU

API id
nextgen

Model identifier

Overview

NextGen is IdonAI’s deepest reasoning model — Pro-class quality available in Quark for hard problems and via the API when keys ship. Large context, multimodal inputs, and mixture-of-experts efficiency for multi-step work.

Specifications

Context

1M tokens

Latency

~20ms

Pricing

From $2.00 / 1M input tokens

Capabilities

  • 182 tokens/sec throughput
  • 1M token context memory
  • Pro-tier reasoning & coding scores
  • Multimodal (text, vision, audio)
  • Mixture-of-experts architecture
  • Advanced security and ethical AI

Benchmarks

MMLU90.6%

Origin 81.4%

GPQA Diamond64.8%

Origin 48.1%

HumanEval89.4%

Origin 78.9%

MATH94.1%

Origin 74.2%

GSM8K96.2%

Origin 91.8%

Context

1M

Ghost bars show the sibling model. Full Origin ↔ NextGen compare lives on the models index.

Ideal use cases

  • Complex multi-step reasoning
  • Enterprise automation
  • Healthcare diagnostics
  • Financial trading analysis
  • Advanced software engineering
  • Scientific research

Call it

Use model id nextgen in the chat completions API. Keys are under development — prototype in Quark today.

Path to production

01

Try in Quark

Prompt NextGen in the assistant before you wire an integration.

Open Quark →
02

Call the API

Use the same model ID in the unified chat completions API.

Read docs →
03

Scale with pricing

Start free on Quark, then pay-as-you-go or enterprise.

View pricing →

Related models

Origin

Our fastest, most responsive model — built for pure speed.

[ Next ]

Run NextGen in Quark.

Same model family behind the assistant and the API. Compare on the models index, then ship.

Try QuarkCompare models