Glimmer Index
Plus $4/mo

DeepSeek V4 Pro

DeepSeek · Released Apr 24, 2026 · Open weights

Open MoE flagship with million-token context for coding and long agent runs

Strongest at
Reading & documents
Better than 49% of the past year's models · one benchmark so far
Price per 1M tokens
$0.96 / $1.91
input / output · OpenRouter
Context window
1M
tokens in · 384K out
Since the last one
1 days
after DeepSeek V4 Pro 0423
Lab announcement ↗Open in the timeline

Capability profile

Scored against 91 models released in the past year. 50 is the average model; further out is better.

This modelMedian of the past year
5090Median of the past yearCoding: 41 · Better than 21%Writing: 54 · Better than 43%Reading & documents: 51 · Better than 49%Reasoning: 48 · Better than 23%Math: 49 · Better than 23%Science: 51 · Better than 9%CodingWritingReadingReasoningMathScience
Coding41Better than 21%
Writing54Better than 43%
Reading & documents51Better than 49% · 1 test
Reasoning48Better than 23%
Math49Better than 23%
Science51Better than 9%

No results yet: Visual reasoning, Computer use

Where these scores come from

Each capability combines the published benchmarks listed here, rescaled so 50 is the reference average and 15 points is one standard deviation. A hollow dot rests on a single benchmark.

Price against capability

The shaded corner holds anything better and cheaper. Prices blend input and output 3 to 1.

5 models from the past year score higher on Reading & documents for less: Gemini 3 Flash Preview, Qwen3.5 9B, DeepSeek V4 Pro 0813 and 2 more.

$0.1$0.3$1$3$10204060GPT-6.1 Terra: 64, $2 per 1MMuse Spark Open: 61, $2 per 1MClaude Haiku 5.5: 39, $0.2 per 1MMistral Large 4: 29, $2.06 per 1MGemini 4 Argon: 71, $4 per 1MGPT-6.1 Sol: 73, $4 per 1MClaude Sonnet 5.5: 67, $4 per 1MGPT-6 Sol: 62, $4 per 1MGPT-6 Luna: 50, $0.2 per 1MClaude Opus 5.5: 68, $8 per 1MGrok 4.7: 56, $3 per 1MDeepSeek V4.1 Flash: 60, $0.21 per 1MGPT-6 Astra: 72, $20 per 1MMuse Spark 1.3: 63, $2 per 1MGemini 3.8 Flash: 55, $1.5 per 1MClaude Fable 5.1: 71, $20 per 1MQwen3.8 Flash: 46, $0.23 per 1MGLM-5.3-Flash: 50, $0.24 per 1MDeepSeek V4 Flash Vision Exp: 33, $0.32 per 1MQwen3.8 27B: 45, $1.12 per 1MGLM-5.3: 47, $2.15 per 1MGemini 3.7 Flash: 60, $1.5 per 1MDeepSeek V4 Pro 0813: 63, $0.99 per 1MGrok 4.6: 53, $3 per 1MMuse Glimmer 30B: 47, $0.52 per 1MMuse Spark 1.2: 54, $2 per 1MQwen3.8 Max: 59, $3 per 1MDeepSeek V4 Flash 0731: 59, $0.15 per 1MClaude Opus 5: 63, $10 per 1MGemini 3.6 Flash: 50, $1.5 per 1MGemini 3.5 Flash Lite: 34, $0.85 per 1MKimi K3: 58, $6 per 1MQwen3.7 Flash: 22, $0.055 per 1MGPT-5.6 Terra: 50, $4.5 per 1MGPT-5.6 Sol: 65, $8 per 1MGPT-5.6 Luna: 44, $0.45 per 1MGrok 4.5: 49, $3 per 1MClaude Sonnet 5: 51, $4 per 1MGLM-5.2: 47, $2.15 per 1MKimi K2.7 Code: 34, $1.71 per 1MClaude Fable 5: 73, $20 per 1MQwen3.7 Plus: 38, $0.7 per 1MClaude Opus 4.8: 55, $10 per 1MQwen3.7 Max: 49, $3.75 per 1MGemini 3.5 Flash: 56, $3.38 per 1MDeepSeek V4 Flash: 31, $0.11 per 1MGPT-5.5: 64, $11.2 per 1MQwen3.6 27B: 30, $1.35 per 1MKimi K2.6: 36, $1.71 per 1MGrok 4.3: 25, $1.56 per 1MGrok Build 0.1: 38, $1.25 per 1MClaude Opus 4.7: 56, $10 per 1MMuse Spark 1.1: 50, $2 per 1MQwen3.6 Plus: 40, $1.12 per 1MGLM-5V-Turbo: 39, $1.9 per 1MGPT-5.4 nano: 36, $0.46 per 1MGPT-5.4 mini: 40, $1.69 per 1MGPT-5.4: 63, $5.62 per 1MQwen3.5 9B: 64, $0.19 per 1MQwen3.5 27B: 26, $0.82 per 1MGemini 3.1 Pro Preview: 69, $4.5 per 1MClaude Sonnet 4.6: 52, $6 per 1MClaude Opus 4.6: 51, $10 per 1MGemini 3 Flash Preview: 65, $1.12 per 1MGPT-5.2 Codex: 50, $4.81 per 1MMistral Large 3: 20, $0.75 per 1MDeepSeek V4 Pro: 51, $1.19 per 1MDeepSeek V4 ProPrice per 1M tokens →↑ Reading & documents$0.1$1$10204060GPT-6.1 Terra: 64, $2 per 1MMuse Spark Open: 61, $2 per 1MClaude Haiku 5.5: 39, $0.2 per 1MMistral Large 4: 29, $2.06 per 1MGemini 4 Argon: 71, $4 per 1MGPT-6.1 Sol: 73, $4 per 1MClaude Sonnet 5.5: 67, $4 per 1MGPT-6 Sol: 62, $4 per 1MGPT-6 Luna: 50, $0.2 per 1MClaude Opus 5.5: 68, $8 per 1MGrok 4.7: 56, $3 per 1MDeepSeek V4.1 Flash: 60, $0.21 per 1MGPT-6 Astra: 72, $20 per 1MMuse Spark 1.3: 63, $2 per 1MGemini 3.8 Flash: 55, $1.5 per 1MClaude Fable 5.1: 71, $20 per 1MQwen3.8 Flash: 46, $0.23 per 1MGLM-5.3-Flash: 50, $0.24 per 1MDeepSeek V4 Flash Vision Exp: 33, $0.32 per 1MQwen3.8 27B: 45, $1.12 per 1MGLM-5.3: 47, $2.15 per 1MGemini 3.7 Flash: 60, $1.5 per 1MDeepSeek V4 Pro 0813: 63, $0.99 per 1MGrok 4.6: 53, $3 per 1MMuse Glimmer 30B: 47, $0.52 per 1MMuse Spark 1.2: 54, $2 per 1MQwen3.8 Max: 59, $3 per 1MDeepSeek V4 Flash 0731: 59, $0.15 per 1MClaude Opus 5: 63, $10 per 1MGemini 3.6 Flash: 50, $1.5 per 1MGemini 3.5 Flash Lite: 34, $0.85 per 1MKimi K3: 58, $6 per 1MQwen3.7 Flash: 22, $0.055 per 1MGPT-5.6 Terra: 50, $4.5 per 1MGPT-5.6 Sol: 65, $8 per 1MGPT-5.6 Luna: 44, $0.45 per 1MGrok 4.5: 49, $3 per 1MClaude Sonnet 5: 51, $4 per 1MGLM-5.2: 47, $2.15 per 1MKimi K2.7 Code: 34, $1.71 per 1MClaude Fable 5: 73, $20 per 1MQwen3.7 Plus: 38, $0.7 per 1MClaude Opus 4.8: 55, $10 per 1MQwen3.7 Max: 49, $3.75 per 1MGemini 3.5 Flash: 56, $3.38 per 1MDeepSeek V4 Flash: 31, $0.11 per 1MGPT-5.5: 64, $11.2 per 1MQwen3.6 27B: 30, $1.35 per 1MKimi K2.6: 36, $1.71 per 1MGrok 4.3: 25, $1.56 per 1MGrok Build 0.1: 38, $1.25 per 1MClaude Opus 4.7: 56, $10 per 1MMuse Spark 1.1: 50, $2 per 1MQwen3.6 Plus: 40, $1.12 per 1MGLM-5V-Turbo: 39, $1.9 per 1MGPT-5.4 nano: 36, $0.46 per 1MGPT-5.4 mini: 40, $1.69 per 1MGPT-5.4: 63, $5.62 per 1MQwen3.5 9B: 64, $0.19 per 1MQwen3.5 27B: 26, $0.82 per 1MGemini 3.1 Pro Preview: 69, $4.5 per 1MClaude Sonnet 4.6: 52, $6 per 1MClaude Opus 4.6: 51, $10 per 1MGemini 3 Flash Preview: 65, $1.12 per 1MGPT-5.2 Codex: 50, $4.81 per 1MMistral Large 3: 20, $0.75 per 1MDeepSeek V4 Pro: 51, $1.19 per 1MDeepSeek V4 ProPrice per 1M tokens →↑ Reading & documents

OpenRouter $0.96 / $1.91 · Cached input $0.08

Since DeepSeek V4 Pro 0423

3 releases in the DeepSeek V4 Pro line.

DeepSeek V4 Pro 0423Apr 23, 20261dDeepSeek V4 ProApr 24, 2026111dDeepSeek V4 Pro 0813Aug 13, 2026
Released1 days later
Context1Msame

A newer release is out: DeepSeek V4 Pro 0813.

ECI STANDING#65 of 270
SCORE149.13147.63–150.87 CI
FRONTIER LAGLaunched 8.5 months behindScores about what GPT-5 did in August 2025.
SCORES LIKE THESE MODELSGPT-5.1OpenAI · Nov 2025149.76
Qwen 3.6 Max (Preview)Alibaba · Apr 2026149.41
Grok 4.3 BetaxAI · Apr 2026148.81
GPT-5.4 MiniOpenAI · Mar 2026148.7
Full Epoch Capabilities Index ranking ↗

Epoch benchmark boards

Best result on each board it has run. The tick marks the leader.

GPQA diamond
91%#36 of 234
OTIS Mock AIME 2024-2025
97%#36 of 216
SWE-Bench verified
78%#6 of 32
DTBench
91%#41 of 179
WeirdML
49%#55 of 141
SimpleQA Verified
47%#36 of 85
CL-bench Life
14%#6 of 14
FrontierMath-Tiers-1-3-v2-Private
45%#57 of 90
6 more boards
Mystery Game Puzzles
17%#60 of 88
ProofBench
16%#58 of 81
Chess Puzzles
19%#67 of 93
Surface Evolver Bench
40%#20 of 26
FrontierCode
18%#40 of 46
FrontierMath-Tier-4-v2-Private
2%#66 of 72

DeepSeek's past year

9 release days in the past year, about one every 41 days. The next is expected around Nov 28, 2026. Forecasts ↗

NovDecJanFebMarAprMayJunJulAugSepOctNov

Specifications

RELEASE DATEApr 24, 2026
WEIGHTSOpen weights
CONTEXT WINDOW1M
MAX OUTPUT384K
REASONINGNot listed
TOOL CALLINGNot listed
INPUTStext
OUTPUTStext

Models.dev record ↗

Specifications and prices from Models.dev; availability and limits may differ by provider. Benchmarks from Epoch AI (CC BY 4.0), LiveBench and the other sources named above.