Back to Calculator
DeepSeek Prover V2 671B
HotDeepSeek's flagship formal theorem prover — 88.9% on MiniF2F-test, SOTA neural theorem proving in Lean 4, built on DeepSeek-V3-Base
Model Specifications
ArchitectureTEXT
Parameters671B
Familydeepseek
VRAM (Q4)335.5GB
Mixture of ExpertsActive inference parameters: 37B.
Same MoE architecture as DeepSeek-V3. Solves 49/658 PutnamBench problems. MIT license.
Estimated Quantization Sizes
| Format | Precision | Est. VRAM | Recommendation |
|---|---|---|---|
| FP16 / BF16 | 16-bit | 1342.0 GB | Uncompressed Base |
| Q8_0High | 8-bit | 671.0 GB | Near Lossless |
| Q6_K | 6-bit | 503.3 GB | Excellent Balance |
| Q4_K_MPopular | 4-bit | 335.5 GB | Standard Use |
Share this Model
Send this model's specs directly to your community.
Similar Models
Related Guides
How much VRAM do you really need?
A complete breakdown of quantization levels and VRAM overhead for running local models.
Best GPUs for Machine Learning in 2026
Comparing NVIDIA and AMD options for the best speed-to-dollar ratio.
GGUF vs EXL2 vs AWQ
Understanding local AI formats and which one to pick for your specific hardware.