29 November 2025
Model
NGen 4 Mini
A smaller NGen 4 reasoning model for high-quality dialogue, multimodal understanding, and scaled product use.
Overview
NGen 4 Mini brings NGen 4 reasoning into a smaller, faster tier. It is designed for high-quality dialogue, creative writing, visual understanding, and general problem solving while keeping serving practical for scaled product use.
Architecture Highlights
- Unified vision-language foundation with early fusion training on multimodal tokens.
- Efficient hybrid architecture for strong throughput and latency characteristics.
- Thinking mode enabled for deeper reasoning on complex logical and mathematical problems.
- Optimized for broad language coverage and culturally nuanced interactions.
Context & Specs
Release
29 November 2025
Context Length
256K (262,144 tokens)
Pricing
Rs 32 input / Rs 56 output per 1M tokens
Best For
Interactive assistants, reasoning products, creative workflows, and scaled multimodal applications.
Serving Notes
NGen 4 Mini operates in thinking mode by default. For production workloads, use dedicated serving stacks such as SGLang, KTransformers, or vLLM and provide enough output length for tasks that need extended reasoning.
Selected release
NGen 4 Mini
1 internal variant
Reasoning
High
Speed
Medium
Input price
Rs 32 / 1M input tokens
Output price
Rs 56 / 1M output tokens
Input
Text, image, audio, video
Output
Text, image, audio, video
Overview
Model capabilities, context, and selection details.
NGen 4 Mini has a 256K context window with a Medium speed tier. Reasoning tier: High. Input support: text, image, audio, video. Output support: text, image, audio, video.
- Context window: 256K (262,144 tokens).
- Speed tier: Medium.
- Reasoning tier: High.
- Input support: text, image, audio, video. Output support: text, image, audio, video.
Context window
256K (262,144 tokens)
Max output
64K (65,536 tokens)
Knowledge cutoff
Jan 2025
Rate limit
1,800 RPM
Pricing
Pricing is based on the selected internal release.
Input
Rs 32 / 1M input tokens
Cached input
Rs 8 / 1M cached input tokens
Output
Rs 56 / 1M output tokens
Model Evaluations
NGen 4 Mini Evaluations
ShadCN benchmark charts for NGen 4 Mini using the latest benchmark sheet you provided.
NGen 4 Mini Reasoning & Knowledge
Instruction following, graduate-level reasoning, contest math, and multilingual knowledge.
NGen 4 Mini Vision & Multimodal
Visual reasoning, embodied reasoning, document understanding, and video reasoning.
Build with TNSA models
Explore model families, snapshots, and developer tools for production AI systems.