Compare LLM Models
Select 2-5 models to compare side by side.
Select models to compare
0 / 5 selected
Top downloads
Select 2-5 models
| Attribute |
sarvamai
|
|---|---|
| Company | sarvamai |
| Release Date | 2026-02-18 |
| Parameters (B) | 106 |
| Architecture | MoE Transformer (MLA) |
| Context Window | 128000 |
| Input Modalities | text |
| Open Source | Yes |
| License | apache-2.0 |
| Score | 9 |
| VRAM (GB) | 112 |
| Compute | 2× NVIDIA A100/H100 80GB GPUs with tensor parallelism; vLLM or SGLang with NVIDIA ModelOpt FP8 (E4M3) quantization |
| Benchmark: AIME-2025 | 88.3 |
| Benchmark: GPQA-Diamond | 78.7 |
| Benchmark: IFEval | 84.8 |
| Benchmark: LiveCodeBench | 71.7 |
| Benchmark: MATH-500 | 98.6 |
| Benchmark: MMLU | 90.6 |
| Benchmark: MMLU-Pro | 81.7 |
| Benchmark: SWE-Bench-Verified | 45 |
| Pricing (Input) | — |
| Pricing (Output) | — |