DeepSeek V5 (Leaked Staging) vs Gemini 4 Pro (Leaked Benchmark)
A direct, empirical evaluation of DeepSeek V5 (Leaked Staging) (DeepSeek AI) against Gemini 4 Pro (Leaked Benchmark) (Google DeepMind).
DeepSeek V5 (Leaked Staging)
Developed by DeepSeek AI · 1.8 Trillion (Sparse MoE) parameters
Autonomous coding agents scoring 81.4% on SWE-bench Verified, zero-telemetry self-hosting, sub-cent 1M-token processing, and formal mathematical proofs.
Gemini 4 Pro (Leaked Benchmark)
Developed by Google DeepMind · Proprietary Omni-Modal Unified Diffusion Transformer parameters
Repository-scale 2M-token refactoring (88.7% DeepSWE v1.1), headless shell execution (95.3% Terminal-Bench 2.1), and multimodal desktop agent autonomy (86.8% OSWorld 2.0).
| Metric | DeepSeek V5 (Leaked Staging) | Gemini 4 Pro (Leaked Benchmark) |
|---|---|---|
| SWE-bench Verified | 81.4% | 88.7% |
| MMLU-Pro | 87.5% | 93.8% |
| MATH-500 | 96.8% | 97.8% |
| Context Window | 1,000,000 tokens | 2,000,000 tokens |
| Input Pricing (per 1M) | $0.12 | $2.25 |
| Output Pricing (per 1M) | $0.48 | $11.25 |