---
library_name: transformers
license: apache-2.0
license_link: https://huggingface.co/gnani/gnani-evon-v3.3/blob/main/LICENSE
pipeline_tag: text-generation
language:
- en
- hi
- bn
- te
- ta
- mr
- gu
- kn
- ml
- pa
- or
tags:
- nemotron
- pytorch
- moe
- mamba
- indic
- long-context
- reasoning
- tool-calling
---

# gnani-evon-v3.3

<div align="center" style="line-height: 1;">
  <img alt="Gnani.ai" src="./gnani_logo.webp" height="32" style="margin-bottom: 8px;"/>
  <br/>
  <a href="https://huggingface.co/gnani/gnani-evon-v3.3/blob/main/LICENSE" style="margin: 2px;">
    <img alt="License" src="https://img.shields.io/badge/License-Apache_2.0-blue.svg" style="display: inline-block; vertical-align: middle;"/>
  </a>
</div>

## Model Overview

**Model Developer:** Gnani.ai

# gnani-evon-v3.3

**Supported languages:** English, Hindi, Bengali, Telugu, Tamil, Marathi, Gujarati,
Kannada, Malayalam, Odia, and Punjabi.

**Release date:** August 2026

gnani-evon-v3.3 is a hybrid Mamba architecture 30B MOE A3B model trained **end-to-end for best-in-class Indic capability** — continued pretraining, supervised fine-tuning, and reinforcement learning
— delivering deep comprehension, extractive QA, and reasoning across Hindi, Bengali,
Telugu, Tamil, Marathi, Gujarati, Kannada, Malayalam, Odia, Punjabi, and English. Developed by Gnani.ai,
each stage was oriented toward native-script Indic language understanding and instruction
following. It supports 128K context length.

**The objective for this release was best-in-class Indic capability.** gnani-evon-v3.3 concentrates on Indic-language comprehension, extraction, and reasoning depth, giving us a strong foundation to build on in upcoming releases.

## Model Architecture

- **Architecture Type:** Mamba2-Transformer Hybrid Mixture of Experts (MoE)
- **Network Architecture:** Nemotron Hybrid MoE (`nemotron_h`)
- **Parameters:** 30B total / ~3.5B active per token
- **Maximum context length:** 131,072 tokens (128K)
- **Precision:** BF16
 

## Key takeaways

- **MILU:** gnani-evon-v3.3 leads **Sarvam-30B** on all **11 of 11** languages at matched scale,
  and **Sarvam-105B** (≈3.5× larger) on **10 of 11**
- We beat **gpt-5.6-luna**, a hosted frontier model, on **MILU Physics (91.32 vs 88.14)** — the benchmark's largest subject at 4,435 items.
- Statistical parity with **gpt-5.4-nano** on Indic comprehension — **79.46 against 79.69** on the nine-language macro, a 0.23-point difference, and ahead on four of those nine languages plus English.
- Indic extractive QA is our strongest competitive position: we lead every open-weight comparator on **XQuAD-in (64.59 F1, +9.50)** and **XorQA-in (41.83 F1, +7.23)**.

### MILU — exact scores by language

**gnani-evon-v3.3** tops this MILU breakdown — ahead of Sarvam-30B on every language and ahead of
Sarvam-105B on 10 of 11.

| Language | gnani-evon-v3.3 | Sarvam-30B | Sarvam-105B |
|---|---:|---:|---:|
| English | **84.06** | 71.13 | 79.73 |
| Kannada | **82.44** | 71.35 | 78.89 |
| Bengali | **82.30** | 69.81 | 78.53 |
| Hindi | **81.99** | 67.08 | 76.95 |
| Telugu | **79.78** | 67.42 | 75.37 |
| Gujarati | **79.05** | 67.43 | 75.13 |
| Tamil | **78.12** | 65.60 | 74.22 |
| Malayalam | **77.00** | 64.38 | 73.87 |
| Odia | 66.94 | 65.83 | **73.70** |
| Marathi | **78.44** | 64.02 | 73.53 |
| Punjabi | **76.02** | 64.63 | 72.94 |
| **Macro mean (11 langs)** | **78.74** | 67.15 | 75.71 |

## Intended Use

gnani-evon-v3.3 is a general-purpose instruction-following, reasoning, and agentic model
intended for:

* English and Indic language chat and assistant applications
* Multilingual RAG systems with long context (up to 128K)
* AI agent systems with tool calling and multi-turn reasoning
* Code and mathematical reasoning
* Instruction following in English and Indic languages

**Not recommended for:** high-stakes decisions, legal/medical/financial advice, or
deployment without safety evaluation on your target languages and domains.

**Deployment geography:** Global

## Input / Output

- **Input:** Text — chat messages (OpenAI-compatible roles: system, user, assistant, tool)
- **Output:** Text — string, optionally with a reasoning trace
- **Maximum input / output size:** 131,072 tokens
- **Supported languages:** English, Hindi, Bengali, Telugu, Tamil, Marathi, Gujarati,
  Kannada, Malayalam, Odia, Punjabi

## Software Integration

- **Runtime Engine(s):** Transformers ≥ 5.3.0, vLLM ≥ 0.12.0, SGLang
- **Supported Hardware:** NVIDIA H100 80GB, NVIDIA H200, NVIDIA A100
- **Operating System(s):** Linux

## Quick Start Guide

### Use it with Transformers

Requires `trust_remote_code=True`

```python
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

model_id = "gnani/gnani-evon-v3.3"

tokenizer = AutoTokenizer.from_pretrained(model_id, trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    trust_remote_code=True,
    device_map="auto",
)

messages = [
    {"role": "user", "content": "Explain photosynthesis in Hindi."},
]

tokenized_chat = tokenizer.apply_chat_template(
    messages,
    tokenize=True,
    add_generation_prompt=True,
    return_tensors="pt",
).to(model.device)

outputs = model.generate(
    tokenized_chat,
    max_new_tokens=1024,
    temperature=0.6,
    top_p=0.95,
    eos_token_id=tokenizer.eos_token_id,
)
print(tokenizer.decode(outputs[0]))
```

Use `temperature=0.6` and `top_p=0.95` for best results across reasoning, tool calling,
and general instruction following.

To disable reasoning traces, pass `enable_thinking=False` to `apply_chat_template`.

### Use it with vLLM

```bash
pip install -U "vllm>=0.12.0"

vllm serve gnani/gnani-evon-v3.3 \
  --served-model-name model \
  --max-model-len 131072 \
  --port 8000 \
  --trust-remote-code \
  --enable-auto-tool-choice \
  --tool-call-parser qwen3_coder \
  --reasoning-parser nano_v3
```

Example client request:

```shell
curl http://localhost:8000/v1/chat/completions \
    -H "Content-Type: application/json" \
    -d '{
        "model": "model",
        "messages": [{"role": "user", "content": "Write a haiku about monsoon rains in Tamil."}],
        "max_tokens": 4096
    }'
```

To disable reasoning with vLLM, add `"chat_template_kwargs": {"enable_thinking": false}`
to the request body.

### SGLang

```bash
pip install "sglang[all]" ninja
```

Start the server:

```bash
python -m sglang.launch_server \
    --model-path gnani/gnani-evon-v3.3 \
    --served-model-name gnani-evon-v3.3 \
    --dtype bfloat16 \
    --context-length 131072 \
    --mem-fraction-static 0.90 \
    --moe-runner-backend triton \
    --host 0.0.0.0 --port 30000
```

`--moe-runner-backend triton` selects the Triton mixture of experts kernels instead of the
CUTLASS ones. The CUTLASS path is compiled on first use and that build needs a CUDA toolchain
whose compiler and headers are the same version, which is not the case in many pip-only
installs. The Triton path needs no such build and starts cleanly. Drop the flag if your CUDA
toolchain is consistent and you want the CUTLASS kernels.

Add `--tp 2` or higher for longer context or higher concurrency. Lower
`--mem-fraction-static` if the server runs out of memory while capturing graphs.

Send a request to the OpenAI-compatible endpoint:

```bash
curl http://localhost:30000/v1/chat/completions \
  -H 'Content-Type: application/json' \
  -d '{
    "model": "gnani-evon-v3.3",
    "messages": [
      {"role": "user", "content": "ಸೌರಶಕ್ತಿಯ ಮೂರು ಉಪಯೋಗಗಳನ್ನು ತಿಳಿಸಿ."}
    ],
    "max_tokens": 2048,
    "temperature": 0.6,
    "top_p": 0.95,
    "chat_template_kwargs": {"enable_thinking": true}
  }'
```

### Recommended Inference Settings

Two request configurations cover most use cases. Choose based on whether the task needs
multi-step reasoning.

**Deterministic single answer** — factual QA, classification, short structured replies
(e.g. multiple-choice, one-line answers). Reasoning off, near-greedy:

```json
{
  "model": "model",
  "messages": [
    {"role": "system", "content": "Respond in the same language and script as the question."},
    {"role": "user", "content": "..."}
  ],
  "temperature": 0.6,
  "top_p": 0.95,
  "repetition_penalty": 1.1,
  "max_tokens": 2048,
  "chat_template_kwargs": { "enable_thinking": false },
  "stream": false
}
```

**Reasoning** — math, physics, multi-step problems where accuracy matters more than
latency. Reasoning on, with `temperature=0.6` and `top_p=0.95`:

```json
{
  "model": "model",
  "messages": [
    {"role": "system", "content": "Respond in the same language and script as the question."},
    {"role": "user", "content": "..."}
  ],
  "temperature": 0.6,
  "top_p": 0.95,
  "repetition_penalty": 1.0,
  "max_tokens": 16384,
  "chat_template_kwargs": { "enable_thinking": true },
  "stream": false
}
```

Notes:

* Non-reasoning mode is faster but slightly less accurate on prompts that require
  multi-step reasoning; enable thinking for those.
* Avoid combining a hard `no_repeat_ngram_size` with `repetition_penalty` on very short answers — legitimate repetition (units, option letters) can be blocked, forcing awkward
  phrasing.
* For strictly native-script output, keep a system instruction that names the target language and script, and verify the response script downstream.

## Training Overview

gnani-evon-v3.3 was built specifically for **Indic capabilities** — comprehension, extraction,
reasoning, and instruction following across English and Indic languages. Every training stage
was oriented toward that goal:

- **Pretraining** on a blended English + Indic corpus to strengthen multilingual
  foundations and long-context Indic representation.
- **Supervised fine-tuning (SFT)** on instruction and capability-focused datasets covering
  QA, summarization, safety, and multi-turn dialogue in native scripts.
- **Reinforcement learning (GRPO)** to sharpen reasoning quality, refusal behavior, and
  task completion on Indic-heavy prompts.

Training ran on **H200 GPU clusters** using NVIDIA's NeMo software stack: **NeMo Curator**, **Megatron Bridge**, and **NeMo RL**.

**Data collection & labeling:** Hybrid — automated, human, and synthetic.

## Evaluation

### Indic Benchmarks

#### Indic Frontier Models Comparison
| Benchmark | Metric | gnani-evon-v3.3 | Sarvam-30B | Sarvam-105B |
|---|---|---:|---:|---:|
| milu | accuracy | **78.74** | 67.15 | 75.71 |
| indicmmlu_pro | accuracy | 68.79 | — | — |
| bharatmath | accuracy | **73.97** | 60.32 | — |
| xquad_in | F1 | **64.59** | 37.86 | 56.65 |
| xorqa_in | F1 | **41.83** | 22.86 | 34.98 |
| flores_in | chrF++ | 34.52 | 18.05 | **34.56** |
| crosssum_in | chrF | 4.67 | 9.66 | **18.01** |
| indicifeval_trans | prompt-strict | 47.45 | 40.76 | **51.79** |
| indicifeval_ground | prompt-strict | 28.66 | **40.69** | 37.88 |
| indicsafe5 | accuracy | **86.50** | 81.26 | — |
 
#### Open Source Models Comparison
| Benchmark | Metric | gnani-evon-v3.3 | Sarvam-30B | Sarvam-105B | gemma-4-26B-A4B | gemma-4-31B |
|---|---|---:|---:|---:|---:|---:|
| milu | accuracy | 78.74 | 67.15 | 75.71 | 82.65 | **85.29** |
| indicmmlu_pro | accuracy | 68.79 | — | — | 70.34 | **79.02** |
| bharatmath | accuracy | 73.97 | 60.32 | — | 76.48 | **78.61** |
| xquad_in | F1 | **64.59** | 37.86 | 56.65 | 55.09 | 42.21 |
| xorqa_in | F1 | **41.83** | 22.86 | 34.98 | 34.47 | 30.16 |
| flores_in | chrF++ | 34.52 | 18.05 | **34.56** | — | — |
| crosssum_in | chrF | 4.67 | 9.66 | 18.01 | 18.85 | **23.09** |
| indicifeval_trans | prompt-strict | 47.45 | 40.76 | 51.79 | 75.20 | **81.12** |
| indicifeval_ground | prompt-strict | 28.66 | 40.69 | **37.88** | 34.73 | 37.79 |
| indicsafe5 | accuracy | 86.50 | 81.26 | — | 88.15 | **89.77** |

### Safety and Security Evaluation

Evaluated on **IndicSafe** — **5,965 rows** across **12 Indic languages**.

**Overall accuracy:** 86.50%

#### Accuracy by Harm Category (%)

| Harm Category | gnani-evon-v3.3 |
|---|---:|
| Harmless Control | 97.93 |
| Tricky Ambiguous | 97.77 |
| Offensive / Hate Speech | 89.80 |
| Harmful Instructions | 85.62 |
| Misinformation / Conspiracy | 83.50 |
| Health Misinformation | 80.41 |
| Religious Hate / Provocation | 80.08 |
| Political Manipulation | 72.49 |
| Gender & Caste Discrimination | 70.26 |

**Key observations:**

- Strongest categories: **Harmless Control** (97.93%) and **Tricky Ambiguous** (97.77%).
- **Misinformation / Conspiracy** (83.50%) is a relative strength among harm categories.
- **Gender & Caste Discrimination** (70.26%) is the hardest category.

### MILU by Language — vs Hosted Models (identical items & prompts)

| Language | gnani-evon-v3.3 | gpt-5.4-nano | DeepSeek-V4-Flash |
|---|---:|---:|---:|
| Hindi | 81.99 | 82.29 | 76.02 |
| English | **84.06** | 83.05 | 78.93 |
| Telugu | **79.78** | 79.52 | 71.15 |
| Bengali | **82.30** | 81.63 | 76.44 |
| Kannada | 82.44 | **82.45** | 76.03 |
| Gujarati | **79.05** | 79.01 | 71.94 |
| Malayalam | **77.00** | 76.93 | 68.85 |
| **Indic macro (9 langs)** | 79.46 | 79.69 | 72.48 |


## Limitations & Ongoing Work

* **Tool use and agentic capabilities** are an active area of development. We are working to
  improve multi-step tool use to match frontier-model capabilities at a similar model size.


## Ethical Considerations

* Evaluate the model on their target languages, domains, and use cases before deployment.
* Implement appropriate safety guardrails for production use.
* Be aware of potential biases in web-scraped and synthetic training data.
* Not use the model for high-stakes decisions without human oversight.
* Ensure any voice or conversational agent built on this model discloses that it is AI.

**Looking ahead:** this model is the Indic-capability foundation for our next plan, where we focus on improving the model further — layering in strong agentic capabilities such as multi-step tool use, planning, and orchestration.

## Citation

```bibtex
@misc{gnani_evon_v33_2026,
  title  = {{gnani-evon-v3.3}: Hybrid {Mamba} {MoE} 30B-A3B Model Trained End-to-End for Best-in-Class Indic Capability},
  author = {{Gnani.ai}},
  year   = {2026},
  url    = {https://huggingface.co/gnani/gnani-evon-v3.3}
}
```