priorsBuilder

DeepSeek · released 2026-04-24

DeepSeek V4 Flash 0420

Measured on 27 benchmarks across 3 reasoning efforts.

General frontier

32.2

#75 of 136 models · best at max

includes estimates: 24% of the mix is measured for this model

Reasoning efforts in your mix

max32.2#204 of 313
high28.9#228 of 313
none14.1#299 of 313

Your mix’s benchmarks

2 of 8 measured

reasoning

Humanity's Last Exam

587 results
  1. #120max of 58734.8%
  2. #144high of 58730.3%
  3. #345none of 5877.8%

Leader: Opus 5.5 · 61.4%

knowledge

AA-Omniscience

526 results
  1. #248high of 526-23.1
  2. #252max of 526-23.9
  3. #332none of 526-44.0

Leader: Opus 5.5 · 46.4

All 27 benchmarks

show 25 more

agentic

AA-AnalystAgent

44 results
  1. #26max of 4425.0%

Leader: Gemini 3.7 Flash · 60.0%

agentic

AA-Briefcase v1.1

224 results
  1. #112high of 2241044
  2. #146max of 224836

Leader: Sonnet 5.5 · 1823

agentic

AutomationBench-AA

219 results
  1. #114high of 21928.5%
  2. #117max of 21928.0%

Leader: Gemini 4 · 77.5%

agentic

EnterpriseOps-Gym-AA

50 results
  1. #30max of 5039.6%

Leader: Fable 5 · 51.1%

agentic

ITBench-AA

45 results
  1. #31max of 4531.5%

Leader: GPT-5.6 Sol · 56.2%

agentic

τ²-Bench Telecom

393 results
  1. #21high of 39395.6%
  2. #27max of 39395.0%
  3. #31none of 39394.4%

Leader: GLM 5.2 · 99.1%

agentic

τ³-Banking

206 results
  1. #63max of 20630.9%
  2. #79high of 20626.2%

Leader: Qwen 3.8 Max · 51.3%

coding

SciCode

226 results
  1. #139max of 22645.3%
  2. #167high of 22640.2%

Leader: Opus 5.5 · 66.9%

coding

Terminal-Bench Hard-AA

387 results
  1. #62high of 38738.6%
  2. #76max of 38735.6%
  3. #91none of 38734.1%

Leader: GPT-5.6 Sol · 65.9%

coding

Terminal-Bench v2.1-AA

238 results
  1. #93max of 23861.8%
  2. #103high of 23856.9%

Leader: Fable 5.1 · 91.4%

coding

Terminal-Bench v4-AA

216 results
  1. #107high of 2163.0%
  2. #112max of 2162.5%

Leader: Sonnet 5.5 · 63.6%

composite

AA Intelligence Index v4.3.2

610 results
  1. #171high of 61024.4
  2. #175max of 61024.2
  3. #240none of 61018.9

Leader: Opus 5.5 · 57.6

cost

AA Intelligence Index Cost per Task

177 results
  1. #37max of 177$0.09
  2. #39high of 177$0.10

Leader: GPT-6 Luna · $0.00

cost

Cost to Run AA Intelligence Index

177 results
  1. #28high of 177$121
  2. #34max of 177$149

Leader: GPT-6 Luna · $10.63

economics

GDP.pdf

210 results
  1. #124max of 21010.8%
  2. #127high of 21010.4%

Leader: Muse Wonder Full (Max) · 32.6%

economics

GDPval-AA v2.1

288 results
  1. #152max of 2881036
  2. #159high of 2881004

Leader: Opus 5.5 · 1866

instruction

IFBench

399 results
  1. #11max of 39979.2%
  2. #40high of 39973.5%
  3. #188none of 39947.2%

Leader: Grok 4.3 · 83.3%

knowledge

AA-Omniscience Accuracy

526 results
  1. #143max of 52636.8%
  2. #151high of 52634.9%
  3. #213none of 52626.0%

Leader: Fable 5.1 · 67.2%

knowledge

AA-Omniscience Non-hallucination

526 results
  1. #383high of 52611.0%
  2. #491none of 5265.5%
  3. #511max of 5263.9%

Leader: MiniCPM5-1B (Non-reasoning) · 99.1%

knowledge

GPQA Diamond

539 results
  1. #78max of 53989.4%
  2. #108high of 53986.7%
  3. #278none of 53971.6%

Leader: GPT-6 Astra · 96.3%

long-context

AA-LCR v1.1

523 results
  1. #162max of 52374.3%
  2. #190high of 52372.0%
  3. #335none of 52341.7%

Leader: Kimi K3 · 88.7%

long-context

MLCR-AA

94 results
  1. #51max of 9415.0%
  2. #74high of 946.7%

Leader: Sonnet 5.5 · 75.0%

reasoning

CritPt

529 results
  1. #137max of 5297.1%
  2. #170high of 5293.4%
  3. #272none of 5290.3%

Leader: Muse Wonder Full (Max) · 32.9%

speed

Latency (Time to First Token)

351 results
  1. #120max of 3511.6s
  2. #129high of 3511.7s

Leader: Gemini 2.5 Flash-Lite · 0.3s

speed

Output Speed (tokens/s)

351 results
  1. #315high of 35142.8
  2. #329max of 35137.8

Leader: Celeris-1 · 1522.5