Verdict
DeepSeek V4 Pro is a solid mid-pack model: #110 of 406 by SIscope score. Strongest at chat quality, weaker at math. Weights are open, so you can run it yourself.
SIscope score
65
#110 of 406
Human-vote rank
#50
of 419 · score 1451
Price per 1M tokens, blended
No price yet
Not listed on OpenRouter
Context window
No data yet
Speed: coming soon
Shape
Each spoke is a category, 0 at the centre, 100 at the edge.
Category scores
Percentiles: 90 means better than about 90% of tested models.
Human-vote score over time
Text arena, recent updates.
Coding votes over time
WebDev arena, recent updates.
Independent test runs
Run by Epoch AI on the same harness for every model. Bars show the share of questions answered correctly.
GPQA Diamond73%
PhD-level science multiple choice questions
Better than 45% of 333 tested models. Run Aug 6, 2026.
OTIS Mock AIME 2024-2547%
Competition maths problems in the style of AIME
Better than 32% of 313 tested models. Run Aug 6, 2026.
Human-vote arenas
People compare two anonymous answers and pick the better one. Scores work like chess ratings.
| Arena | Rank | Score | Range (95%) | Votes | Updated |
|---|---|---|---|---|---|
| Text arena Everyday chat, blind human votes | #50 of 419 | 1451 | 1447–1455 | 57,572 | Oct 28, 2026 |
| WebDev arena Building web apps, blind human votes | #79 of 152 | 1447 | 1441–1454 | 13,931 | Oct 28, 2026 |
| Vision arena Questions about images, blind human votes | Not on this board yet | ||||
| Agent arena Real multi-step agent sessions | Not on this board yet | ||||
Source: Arena (formerly LMArena) leaderboard dataset, CC BY 4.0. Fetched Nov 4, 2026.
Price
No public list price yet. We only show prices from the OpenRouter models API, and this model isn't listed there under a matching name.