Is DeepSeek V4 Pro Preview SOTA in Reasoning?#72 in ReasoningRanked #72 in Reasoning.
How it scores here
DeepSeek V4 Pro Preview in Reasoning
Primary benchmarks
- LMArena Text (style-controlled, hard prompts): 1473 (CI 1469–1478) on Nov 19, 2026 (#90 of 263).
On the category's weighted score it is 42.9 points behind #1 Gemini 4 Carbon and 0.6 behind the next model up, GPT-5.6 Luna.
Nearby on the board
| # | Model | Category score |
|---|---|---|
| 69 | Gemma 4 31B | 58.8 |
| 70 | Gemini 3.5 Flash Lite | 58.2 |
| 71 | GPT-5.6 Luna | 57.6 |
| 72 | DeepSeek V4 Pro Preview | 57.1 |
| 73 | GPT-5.3 Chat | 56.5 |
| 74 | Qwen3 7 Plus | 55.9 |
| 75 | Mimo V2.6 Flash | 55.3 |
Standings
Reasoning
| # | Model | State |
|---|---|---|
| 1 | Gemini 4 Carbon | Co-SOTA |
| 1 | Gemini 4 Argon | Co-SOTA |
| 1 | Gemini 4 Neon | Co-SOTA |
| 2 | Muse Wonder | Near-SOTA |
| 3 | Claude Opus 4.6 | Near-SOTA |
| 4 | Claude Opus 5.5 | Near-SOTA |
| 5 | Claude Fable 5 | Near-SOTA |
| 6 | Claude Opus 4.7 | Was SOTA |
| 7 | Claude Fable 5.1 | Was SOTA |
| 8 | Muse Spark 1.3 | #8 in this category |
| 9 | Gemini 3.8 Flash | #9 in this category |
| 10 | Claude Fable 5.2 | #10 in this category |