DeepSeek V4 Pro
deepseek-thinking
Open MoE flagship with million-token context for coding and long agent runs
ReasoningTool CallingStructured OutputOpen Weights
Context Window
1M
Max Output
384K
Temperature
Yes
Open Weights
Yes
Knowledge Cutoff
2025-05
Released
2026-04-24
Last Updated
2026-04-24
Modalities
Input:Text
→Output:Text
Benchmarks
| Benchmark | Score | Metric | Harness |
|---|---|---|---|
| SWE-Bench Verified | 80.6 | resolved | - |
| Artificial Analysis Coding Agent Index | 50.1 | average pass@1 | Claude Code |
| SWE-Atlas Codebase QnA | 67.8 | pass@1 | Claude Code |
| SWE-Bench Pro | 18 | pass@1 | Claude Code |
| Terminal-Bench | 64.7 | pass@1 | Claude Code |
Available Providers (10)
| Provider | Input /1M | Output /1M | Cache Read /1M | Cache Write /1M | Reasoning | Status |
|---|---|---|---|---|---|---|
| Rp 7.845 | Rp 15.690 | Rp 65 | — | toggle, effort | — | |
| Rp 28.854 | Rp 57.708 | Rp 2.435 | — | toggle, effort | — | |
| — | — | — | — | ✓ | — | |
| Rp 7.845 | Rp 15.690 | Rp 66 | — | toggle, effort | — | |
| Rp 7.845 | Rp 15.690 | Rp 66 | — | ✓ | — | |
| Rp 10.099 | Rp 20.198 | Rp 66 | — | ✓ | — | |
| Rp 7.845 | Rp 15.690 | Rp 66 | — | toggle, effort | — | |
| Rp 7.845 | Rp 15.690 | Rp 66 | — | toggle, effort | — | |
| Rp 31.379 | Rp 62.758 | — | — | toggle | — | |
| Rp 19.838 | Rp 39.675 | Rp 1.984 | — | toggle, effort | — |