Model record
DeepSeek R1 8B
DeepSeek · MIT · verified
What this model is
The architecture is dense. Every parameter is part of the active path, so memory and speed scale with the full model rather than a smaller expert subset.
Memory and quantization
Decimal GB, rounded from the Ollama tag page. Runtime memory still needs KV cache and overhead.
| Quantization | Published files | 8K estimate |
|---|---|---|
| Q4_K_M | 4.90 GB | 7.47 GB |
| Q8_0 | 8.50 GB | 11.07 GB |
Estimate formula: published weight files + 1.0737 GB per 8K tokens × context / 8K + 1.5 GB runtime overhead. Sizes are decimal GB, not GiB. At 8K context, the smallest listed build is about 7.5 GB before OS and application reserve.
Conservative FP16 estimate from the Llama 3.1 8B GQA configuration used by this distillation; not a runtime measurement.
KV method source: https://ollama.com/library/deepseek-r1
Speed heuristic
Speed depends on quantization, backend, and hardware; no single measured value is reported
Active path: 8B parameters. This is a relative speed signal only; it is not a tokens/sec benchmark.
Runtimes
- Ollama
- llama.cpp
ollama run deepseek-r1:8bLicense restrictions
The DeepSeek release is MIT, but the Llama 3.1 base-model terms also apply to this derivative.
This is a plain-language summary, not legal advice. Read the linked license before commercial deployment or redistribution.
Known issues
- Reasoning output can be long and may need more generation time
- This is a Llama 3.1 distillation, not the full DeepSeek R1 MoE model
Sources
- https://ollama.com/library/deepseek-r1
- https://huggingface.co/deepseek-ai/DeepSeek-R1-Distill-Llama-8B
This page is an independent compatibility guide, not official documentation. Verify the primary source before downloading or deploying a model.