DeepSeek V4 Pro benchmarks
DeepSeek V4 Pro has 11 published scores in this dataset, from a DeepSeek release dated 2026-08-13. Its composite index is 91.0, which places it 6th of 22 ranked models and 2nd among 15 open-weight models. Its strongest area is knowledge (index 100.0) and its weakest is reasoning (index 86.7). Its best placing is on LiveCodeBench, where its 93.5 ranks 1st of 7 models with a published score.
Category profile
Index 0–100: each score relative to the best published score on that test.
Model facts
- Lab
- DeepSeek
- Type
- Open weights
- Released
- 2026-08-13
- Parameters
- 1.6T total, 49B active (MoE)
- Context
- 1M
- License
- MIT
DeepSeek V4 Pro hardware requirements and cost
Even at 4-bit it needs about 1.1 TB (8× H200 (1,128 GB)), well beyond a home workstation, so most people run it on rented multi-GPU servers. Full BF16 precision needs about 3.7 TB (Multiple nodes (over 1.1 TB)).
| Precision | Weights | With headroom | Smallest setup |
|---|---|---|---|
| BF16 (full precision) | 3.2 TB | 3.7 TB | Multiple nodes (over 1.1 TB) |
| FP8 / 8-bit | 1.6 TB | 1.8 TB | Multiple nodes (over 1.1 TB) |
| 4-bit (GGUF Q4_K_M) | 960 GB | 1.1 TB | 8× H200 (1,128 GB) |
Estimate: parameters × bytes per parameter, plus 15% for the KV cache and runtime at modest context lengths. Long contexts and large batches need more.
Lighter or stronger open models
Open models ranked just above and below on the composite index, with their 4-bit memory needs.
- Kimi K3.1 Pro94.4—
- Kimi K390.61.9 TBcompare
- Tencent Hy4 Preview89.9531 GB
- Muse Spark Open88.0563 GB
DeepSeek V4 Pro speed and pricing
Measured figures link to their source; estimates are labelled.
- API output speed
- 94 tok/sArtificial Analysis median · Artificial Analysis
- Time to first token
- 1.00 sOpenRouter model page provider stats
- API price per 1M tokens
- $1.32 in · $3.96 outlab list price · DeepSeek API pricing docs
Score by test
Rank is among every model with a published score on that test.
Nearby in the composite ranking
Models with the closest composite index.
Head-to-head comparisons
Side-by-side pages built from the tests both models report.
Sources
Scores are published results from model cards, lab reports, and public leaderboards, each linked to its source. Settings such as thinking mode or test version differ between sources and are shown next to each score. OpenLLM Buddy catalog models are uncensored fine-tunes; their scores are those of the official base model. Data collected 2026-10-30. How the index works