Sarvam 1 — Hardware Requirements & GPU Compatibility
Chat
Sarvam 1 is a 2.5B-parameter open language model from sarvamai. It supports a context window of up to 8,192 tokens. At BF16 it needs about 5.59 GB of VRAM — see which GPUs and Macs can run it below.
6.2K downloads
138 likesSpecifications
- Publisher
- sarvamai
- Parameters
- 2.5B
- Architecture
- LlamaForCausalLM
- Context Length
- 8,192 tokens
- Vocabulary Size
- 68,096
- Release Date
- 2024-11-08
Get Started
HuggingFace
How Much VRAM Does Sarvam 1 Need?
Select a quantization to see compatible GPUs below.
| Quantization | Bits | VRAM | + Context | File Size | Quality |
|---|---|---|---|---|---|
| BF16 | 16.00 | 5.6 GB | 6.3 GB | 5.05 GB | Brain floating point 16 — preferred for training |
Which GPUs Can Run Sarvam 1?
BF16 · 5.6 GBSarvam 1 (BF16) requires 5.6 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 8+ GB is recommended. Using the full 8K context window can add up to 0.7 GB, bringing total usage to 6.3 GB. 35 GPUs can run it, including NVIDIA GeForce RTX 5090, NVIDIA GeForce RTX 3090 Ti, NVIDIA GeForce RTX 3070 Ti.
Runs great
— Plenty of headroom
Which Devices Can Run Sarvam 1?
BF16 · 5.6 GB33 devices with unified memory can run Sarvam 1, including NVIDIA DGX H100, NVIDIA DGX A100 640GB, MacBook Air 13" M3 (8 GB).
Runs great
— Plenty of headroomDecent
— Enough memory, may be tightRelated Models
Frequently Asked Questions
- How much VRAM does Sarvam 1 need?
Sarvam 1 requires 5.6 GB of VRAM at BF16. Full 8K context adds up to 0.7 GB (6.3 GB total).
VRAM = Weights + KV Cache + Overhead
Weights = 2.5B × 16 bits ÷ 8 = 5.1 GB
KV Cache + Overhead ≈ 0.5 GB (at 2K context + ~0.3 GB framework)
KV Cache + Overhead ≈ 1.2 GB (at full 8K context)
VRAM usage by quantization
BF165.6 GBBF16 + full context6.3 GB- Can I run Sarvam 1 on a Mac?
Sarvam 1 requires at least 5.6 GB at BF16, which exceeds the unified memory of most consumer Macs. You would need a Mac Studio or Mac Pro with a high-memory configuration.
- Can I run Sarvam 1 locally?
Yes — Sarvam 1 can run locally on consumer hardware. At BF16 quantization it needs 5.6 GB of VRAM. Popular tools include Ollama, LM Studio, and llama.cpp.
- How fast is Sarvam 1?
At BF16, Sarvam 1 can reach ~522 tok/s on AMD Instinct MI300X. On NVIDIA GeForce RTX 4090: ~117 tok/s. Speed depends mainly on GPU memory bandwidth. Real-world results typically within ±20%.
tok/s = (bandwidth GB/s ÷ model GB) × efficiency
Example: AMD Instinct MI300X → 5300 ÷ 5.6 × 0.55 = ~522 tok/s
Estimated speed at BF16 (5.6 GB)
~522 tok/s~117 tok/s~390 tok/s~322 tok/sReal-world results typically within ±20%. Speed depends on batch size, quantization kernel, and software stack.
- What's the download size of Sarvam 1?
At BF16, the download is about 5.05 GB.
- Which GPUs can run Sarvam 1?
35 consumer GPUs can run Sarvam 1 at BF16 (5.6 GB). Top options include AMD Radeon RX 6700 XT, AMD Radeon RX 6800, AMD Radeon RX 6800 XT, AMD Radeon RX 7600. 28 GPUs have plenty of headroom for comfortable inference.
- Which devices can run Sarvam 1?
33 devices with unified memory can run Sarvam 1 at BF16 (5.6 GB), including Mac Mini M4 (16 GB), Mac Mini M4 (32 GB), Mac Mini M4 Pro (24 GB), Mac Mini M4 Pro (48 GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.