December 2025
Chat Model
NGen 3.9 Max
The strongest production snapshot in the NGen 3.9 family, built for demanding enterprise reasoning.
Overview
NGen 3.9 Max Stable V3 is the flagship production model in the NGen 3.9 family. It is tuned for reliability, maximum 3.9-series performance, and high-confidence reasoning across production workloads.
Context & Specs
Context Length
262,144 tokens
Model Priority
Stable production flagship
Best For
Reasoning-heavy enterprise workflows, knowledge work, coding, and multimodal analysis.
Pricing
Rs 0.83 input / Rs 1.25 output per 1K tokens
Key Capabilities
- Maximum production stability and reliability for the NGen 3.9 series.
- Strong performance in reasoning, knowledge, and coding evaluations.
- Advanced multimodal understanding and document/vision coverage.
- Enterprise-grade safety and alignment features.
Snapshots
NGen 3.9 Max Stable V3
Production flagship snapshot.
NGen 3.9 Max V2 128K Preview
Long-context preview snapshot.
Availability
Served from TNSA’s Indian regions.
India South
in-south
India Central
in-central
US East-1
Not available
NGen 3 Series Reasoning & Knowledge
Zero-shot performance of the NGen 3.9 and 3.5 series.
- NGen 3.9 Max
- NGen 3.9 Pro
- NGen 3.5 Max
- NGen 3.5 Pro
- OpenAI o3-mini
- OpenAI o4-mini
- DeepSeek R1 0528
Advanced Text & Reasoning
Comparing NGen 4 and NGen 3 series against Qwen3, DeepSeek, and Llama 4.
- NGen 4 Pro
- NGen 4 Mini
- NGen 3.9 Max
- NGen 3.9 Pro
- NGen 3.5 Max
- Qwen3 30B A3B
- DeepSeek R1 0528
- Llama 4 Maverick
Logic & Mathematical Reasoning
Frontier performance on AIME 25 and HMMT 25 benchmarks.
- NGen 4 Pro
- NGen 4 Mini
- NGen 3.9 Max
- NGen 3.5 Max
- Qwen3 4B (2507)
- Qwen3 4B Think
Advanced Engineering & Coding
Comparative analysis on LiveCodeBench and CFEval.
- NGen 4 Pro
- NGen 4 Mini
- NGen 3.9 Max
- NGen 3.5 Max
- DeepSeek R1 0528
- Llama 4 Maverick
Text, Reasoning & Coding Benchmarks
| Category | Benchmark | Score | Industry Avg. |
|---|---|---|---|
| Knowledge | MMLU-Pro | 81.8 | 81.2 |
| Knowledge | MMLU-Redux | 93.1 | - |
| Reasoning | GPQA | 73.8 | 68.4 |
| Reasoning | AIME 25 | 88.3 | 70.9 |
| Coding | LiveCodeBench | 61.2 | 57.4 |
| Alignment | Arena-Hard v2 | 40.9 | 36.3 |
Vision & Multimodal Benchmarks
| Category | Benchmark | Score | Industry Avg. |
|---|---|---|---|
| STEM | MMMU (Val) | 75.6 | 75.8 |
| Math | MathVista | 83.2 | 81.4 |
| OCR | DocVQA | 95.6 | 95.3 |
| Video | MVBench | 74.7 | 69 |
Build with TNSA models
Explore model families, snapshots, and developer tools for production AI systems.