Skip to content

DeepSeek V4 Flash 0731

ModelOpen sourceActive
by DeepSeek · family “deepseek-v4-flash”

DeepSeek V4 Flash 0731 is a sparse mixture-of-experts model from DeepSeek, with 13B active parameters out of 284B total. This re-post-trained revision is suited for coding, reasoning, and agent workflows.... (description from the OpenRouter listing)

in: textout: textReasoningTool useStructured output
Epoch Capabilities Index
154.5 #31 of 268
90% CI 152.1 – 156.5
Listing ↗
Released
31 Jul 2026
Context window
1.31M tokens
Max output
944K tokens
Input $ / 1M tokens
$0.018
Output $ / 1M tokens
$0.32
Cached input $ / 1M
$0.018
Each benchmark shown separately with its own source

Benchmark results (12)

All benchmarks
BenchmarkDomainScorevs best recordedSettingRunSource
LMCAagents41.7%
61%
max—External ↗
DTBenchreasoning90.9%
92%
max—External ↗
Mystery Game Puzzlesgames34.0% ±4.8
40%
max5 Aug 2026Epoch ↗
FrontierMath-Tiers-1-3-v2-Privatemath57.5% ±2.9
61%
max2 Aug 2026Eval log ↗
FrontierMath-Tier-4-v2-Privatemath24.4% ±6.8
25%
max2 Aug 2026Eval log ↗
Chess Puzzlesgames33.0% ±4.7
46%
max2 Aug 2026Epoch ↗
SimpleQA Verifiedknowledge33.6% ±1.5
44%
max27 Aug 2026Eval log ↗
ARC-AGI-2reasoning61.4%
65%
max—External ↗
WeirdMLcoding63.0% ±0.0
67%
max—External ↗
OTIS Mock AIME 2024-2025math94.4% ±2.1
94%
max2 Aug 2026Epoch ↗
GPQA diamondscience91.0% ±1.8
95%
max2 Aug 2026Epoch ↗
ARC-AGIreasoning89.0%
90%
max—External ↗

Source: Epoch AI Benchmarking Hub (CC BY 4.0). “External” rows are leaderboard results Epoch collects from third parties. Best reported setting per benchmark is shown.

API list price over time

Price history

$ per 1M tokens

4 recorded prices on 3 Aug 2026 (OpenRouter listing + Internet Archive snapshots); re-read on every data refresh, most recently 6h ago. Steps show when the price changed.

Events