Sarvam 105b needs about 65.3 GB of memory at Q4_K_M (a 64.2 GB file) with an 8K context window, which is more than a 48 GB GPU holds. Longer context needs more: at 32K tokens the KV cache alone adds about 2.3 GB.
Sarvam 105b – a 106B‑parameter MoE model in GGUF format
Sarvam 105b is a 106B‑parameter Mixture‑of‑Experts model based on the deepseek2 architecture. It is optimized for complex reasoning, agentic tasks, mathematics, and coding. The model is released under the Apache‑2.0 license.
Good at
- Complex reasoning
- Agentic tasks
- Mathematics and coding
Limits
- Requires large VRAM – 65.3 GB at 8 k context
- No vision support
- License is Apache‑2.0
Best for
- High‑complexity reasoning workflows
- Agentic task automation
- Advanced coding assistance
Who it’s for. Developers and researchers needing a powerful reasoning engine for large‑context tasks. The model is suitable for applications that can accommodate its high memory footprint.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q4_K_M | 64.2 GB (9 parts) | 65.3 GB | Too big | Get file → |
The only available quant is Q4_K_M, 64.2 GB. It needs about 65.3 GB of VRAM at 8 k context. For lower memory budgets, the model is not supported in the current release.
What GPU do you need for Sarvam 105b?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: no quant fits at an 8K context.
- 12 GB VRAM: no quant fits at an 8K context.
- 16 GB VRAM: no quant fits at an 8K context.
- 24 GB VRAM: no quant fits at an 8K context.
- 32 GB VRAM: no quant fits at an 8K context.
- 48 GB VRAM: no quant fits at an 8K context.
Which Sarvam 105b quant should you choose?
Q4_K_M is the usual starting point for Sarvam 105b: it balances file size and quality well. Here it is a 64.2 GB file.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Sarvam 105b
How much VRAM does Sarvam 105b need?
About 65.3 GB at Q4_K_M with an 8K context window. That is the 64.2 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Sarvam 105b on an 8 GB GPU?
Not at an 8K context. The smallest file needs about 65.3 GB, so you would need a larger GPU or CPU offloading.
What is the file size of Sarvam 105b GGUF?
It depends on the quant: Q4_K_M is 64.2 GB.
What context length does Sarvam 105b support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.6 GB at 8K and 2.3 GB at 32K.
Where can I download Sarvam 105b GGUF?
The files are on Hugging Face in the sarvamai/sarvam-105b-gguf repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Sarvam 105b use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run Sarvam 105b on 8 GB VRAM?
No, the model requires approximately 65.3 GB of VRAM at 8 k context.
Does Sarvam 105b support vision tasks?
No, the model does not include vision capabilities.