DeepSeek V4 Pro snapshot with million-token context and support for thinking and non-thinking modes
Specification
most-agreed values
Context
1M
Max output
384K
Released
2026-08-12
Knowledge cutoff
—
Retires
—
Open weights
yes
Input
text
Output
text
Price
US dollars per million tokens · most-agreed
Input
$1.3359
Output
$4.0077
Cache read
$0.0445
Quality
8 benchmarks · 2 sources
Composite
91st
percentile of 334 scored models
Rank
30 / 334
±0.08 sd
Evidence
8 × 2
3 or more benchmarks, from 2 or more sources
Effort range
—
one configuration only
reasoning97th3/4 bench
math83rd3/7 bench
knowledge78th1/1 bench
general71st1/1 bench
Every score
one row per source and configuration — nothing averaged away
| Benchmark | Score | Configuration | Source | Run |
|---|---|---|---|---|
| Chess Puzzles | 47.0 ±5.0 | max | Epoch AI | 2026-08-18 |
| Mystery Game Puzzles | 43.0 ±5.0 | max | Epoch AI | 2026-08-19 |
| OTIS Mock AIME 2024-2025 | 98.6 ±1.0 | max | Epoch AI | 2026-08-18 |
| GPQA diamond | 91.7 ±1.5 | max | Epoch AI | 2026-08-18 |
| FrontierMath-Tiers-1-3-v2 | 64.6 ±2.8 | max | Epoch AI | 2026-08-19 |
| SimpleQA Verified | 52.9 ±1.6 | max | Epoch AI | 2026-08-27 |
| LiveBench | 78.2 | default · 23 runs | LiveBench | 2026-06-25 |
| FrontierMath-Tier-4-v2 | 26.8 ±7.0 | max | Epoch AI | 2026-08-19 |
The leading figure for a benchmark is the median across its configurations, so one heroic high-effort run cannot set the number.
Available from 1 host
| Host | In $/M | Out $/M | Cache rd | Cache wr | Context | Output | Retires |
|---|---|---|---|---|---|---|---|
| volcengine | $1.3359 | $4.0077 | $0.0445 | — | 1M | 384K | — |
More from DeepSeek
most-hosted first