Get started
Models
All language and reasoning pipelines run on the TNSA-LM engine. NGen 4.3 is the latest family; choose a model based on the intelligence, context length and latency your workload needs.
NGen 4.3 — latest
Frontier intelligence with a million-token context window and Low, Medium, High and Extended reasoning levels. Call it with /infer or the OpenAI-compatible endpoint.
| Model | Description | Context |
|---|---|---|
| NGen-4.3-ProNew | Newest flagship for long-horizon reasoning, software engineering, computer use and multimodal agentic work. Up to 248,320 output tokens. | 1M |
NGen 4.1
Next-generation models with improved reasoning, vision and long-context performance.
| Model | Description | Context |
|---|---|---|
| NGen-4.1-Pro | Most capable NGen 4.1 model across knowledge, reasoning, engineering and agent tasks. | 256K |
| NGen-4.1-LiteNew | Next-generation Lite tier with much stronger knowledge and reasoning at a fast, low price. | 256K |
| NGen-4.1Deprecated | Next-generation flagship for knowledge, reasoning and engineering. | 256K |
| NGen-4.1-MiniDeprecated | Mini model with strong vision, OCR and long-context reasoning. | 256K |
Deprecations
NGen-4
The established flagship family for coding, reasoning and production assistants.
| Model | Description | Context |
|---|---|---|
| NGen-4 | Flagship NGen-4 for structured step-by-step reasoning and multimodal problem solving. | 256K |
| NGen-4 Mini | Compact NGen-4 for low-latency assistants and high-volume routing. | 256K |
| NGen-4 Lite | Balanced cost, speed and reasoning for everyday workloads. | 512K |
| NGen-4 Flash | High-speed, non-reasoning variant for realtime UX and short-turn chat. | 128K |
| NGen-4 Blaze | Lowest-reasoning, highest-speed variant for instant replies and routing. | 128K |
NGen 3.9 & 3.1
Earlier generations, still available at lower rates. NGen3.1-Pro adds image understanding.
| Model | Description | Context |
|---|---|---|
| NGen3.9-Pro | Reasoning engine with deep logic, code and math. | 40K |
| NGen3.9-Lite | Lightweight, low-latency model for dialogue and classification. | 40K |
| NGen3.9-Max-v2-128K | High-capacity model for large documents and long logs. | 128K |
| NGen3.9-Max-Stable-v3 | Stable checkpoint for audited, enterprise responses. | 128K |
| NGen3.1-Pro | Vision-tuned reasoning model that handles image inputs. | 40K |
Image models (IGen-1)
Generate and edit images via /generate-image.
| Model | Description | Resolution |
|---|---|---|
| IGen-1 | Full-quality image generation and editing. | 1024px |
| IGen-1 Fast | Faster image branch for practical generation workflows. | — |
| IGen-1 Nano | Lowest-cost image model for drafts and fast iterations. | 512px |
Audio models
Transcribe, synthesize and run conversational voice via the audio endpoints, with support for 13+ Indian languages.
| Model | Description | Context |
|---|---|---|
| AudioTextSpeech-Pro | Conversational voice: transcription + reasoning + synthesis in one call. | — |
| AudioTextSpeech-1 | Standalone transcription and speech synthesis. | — |
| NGen-Realtime-Omni | Low-latency model for live streaming audio and speech. | 8K |
In the Platform playground
Choosing a model
Start with NGen-4.3-Pro for the strongest reasoning, coding and long documents. For fast, high-volume work use NGen-4.1-Lite or NGen-4 Mini, and NGen-4 Flash / Blaze when latency matters most. See Pricing for per-model rates.