DeepSeekFree TierMoE SparseZero Data Retention

V4 Flash-0731

Name: ARMES
Availability: InStock
Author: ARMES

Run V4 Flash-0731

Included on FREE plan

Context Window

1M (384K max output)

Input Token Price

$0.14

per 1,000,000 tokens

Output Token Price

$0.28

per 1,000,000 tokens

Speed Rating

Fast

Verified Frontier Benchmarks

Empirical Evaluation Results

Evaluated across standard and verified harnesses

AA Intelligence Index

reasoning

34.3

AA Coding Index

coding

69.1

AutomationBench

agentic

54.0%

AA Briefcase

strategy

1259.3 Elo

Critpt

reasoning

16.6%

Gdp Pdf

strategy

11.0%

AA GDPval 2.0

strategy

1467.8 Elo

GPQA Diamond

reasoning

90.8%

Humanity's Last Exam (HLE)

reasoning

38.6%

Lcr (1.1)

reasoning

79.7%

Mlcr

reasoning

13.3%

Omniscience

science

-14.3

SciCode

science

50.3%

Tau3 Banking

finance

39.4%

Terminal-Bench 2.1

coding

78.7%

Terminal-Bench 4.0

coding

12.1%

Coding Agent Index (1.4)

agentic

49.8%

Coding Agent Index (1.5)

agentic

38.7%

Openrouter GPQA Diamond

reasoning

85.1%

Openrouter Tau2 Bench Airline

finance

72.8%

Eqbench Creative Writing (3)

voice

1438.4 Elo

Eqbench Judgemark (4)

voice

30.7

Eqbench Longform Writing (v1.11)

voice

61.2

Frontiermath Tiers 1 3 (v2)

math

57.5%

FrontierMath Tier 4

math

24.4%

SimpleQA

voice

33.6%

Chess Puzzles

reasoning

33.0%

Mystery Game Puzzles

reasoning

34.0%

Epoch GPQA Diamond

reasoning

91.0%

Toolathlon Verified

agentic

70.7%

Context Arena MRCR V2 (8-needle)

agentic

77.9%

AA Agentic Index

agentic

41.7

TAU-Bench Banking

agentic

39.4%

Briefcase

strategy

1259.3 Elo

Gdpval (2)

strategy

1467.8 Elo

TAU-3 Banking

agentic

39.4%

TAU-2 Airline

agentic

72.8%

Briefcase v1.1

strategy

1259.3 Elo

GDPval-AA v2.1

strategy

1467.8 Elo

AA-LCR v1.1

reasoning

79.7%

AA-Omniscience Index

reasoning

-14.3

SimpleQA Verified

voice

33.6%

Creative Writing v3

voice

1438.4 Elo

Arc Agi (1)

reasoning

89.0%

Arc Agi (2)

reasoning

61.4%

Architectural Profile & Capabilities

Terminal-Bench 2.1: 82.7%, CyberGym: 76.7%, Toolathlon-Verified: 70.3%, DSBench-FullStack: 68.7%, DSBench-Hard: 59.6%, DeepSWE: 54.4%, NL2Repo: 54.2%, Agents' Last Exam: 25.2%, AutomationBench Public: 25.1%, Artificial Analysis Intelligence Index: 50 | The official GA release of V4 Flash, superseding the April preview (released Jul 31, 2026). **Same architecture and parameter count — every gain is from re-post-training, not a new model.** Outperforms V4 Pro Preview on 7 of 9 published agent benchmarks despite activating under a third of Pro's parameters (13B vs 49B). Thinking on by default with three reasoning modes (non-think, high, max). Natively supports Responses API with Codex-style agent harness adaptation. DSpark speculative decoding module attached for faster inference. 2,500 concurrent request support (5× V4 Pro). MIT licensed open weights on Hugging Face. **Best for:** high-throughput agentic coding, terminal/shell agent loops, multi-step tool orchestration, security research, data science full-stack tasks, cost-sensitive automation pipelines, and any workload where the ~90× output-price advantage over frontier models matters more than the 2–15 pt capability gap. **Caveats: all benchmark scores are DeepSeek's own internal runs using DeepSeek Harness (unreleased) at max effort — no independent third-party verification at time of writing. NIST CAISI previously found V4 family real-world capability lagging DeepSeek's self-reported figures by ~8 months. Still behind Opus 4.8 on every published benchmark. No system card or safety framework published. Peak/off-peak pricing (2× during Beijing business hours) took effect Aug 16, 2026 — output rose from $0.28 to $1.32 peak / $0.66 off-peak.**

Architecture Type: Mixture of Experts (MoE)

Sparse Mixture-of-Experts routing: activates only a subset of parameter experts per token, delivering frontier-level intelligence with exceptional efficiency.

Recommended Workloads & Primary Use Cases

high-throughput agentic coding

terminal/shell agent loops

multi-step tool orchestration

security research

data science full-stack tasks

cost-sensitive automation pipelines

ARMES Zero Data Retention

Calls to V4 Flash-0731 are routed through strict Zero Data Retention inference channels. Prompts and outputs are never stored, indexed, or monitored by DeepSeek.

View Privacy Architecture

More Models from DeepSeek

Compare other engines available in this laboratory.

View Roster (7)

V4 Pro 0813Eco Tier

Terminal-Bench 2.1: 87.9%, DeepSWE: 62.7%, CyberGym: 83.3%, AutomationBench: 31.8%, Toolathlon-Verified: 74.1%, HLE w/ tools: 60.0%, GPQA Diamond: 90.1%, Artificial Analysis Intelligence Index: 53.2, AA Coding Index: 68.8, AA Agentic Index: 49.6 | The official GA release of V4 Pro, superseding the April preview (released Aug 13, 2026).

1M ctx$1.32/M

V4 ProEco Tier

V4 Pro is a high-performance foundation model engineered by DeepSeek, accessible with Zero Data Retention on ARMES.

128K ctx$1.74/M

V4.1 FlashFree Tier

DeepSeek-V4.1-Flash is DeepSeek's first Causal Encoder-Decoder (CED) frontier model: a 552B-parameter multimodal sparse MoE that activates only 8B parameters during prefill and 16B during decode, with a 1M-token context window, native image understanding, and aggressive FP4/compressed KV caching. It tops agentic terminal and software-engineering boards (90.6 on Terminal-Bench 2.1, 74.2 on DeepSWE v1.1 at max effort) at roughly $0.15/$0.60 per 1M tokens, and supersedes V4-Pro in production traffic.

1.05M ctx$0.30/M

Architectural Profile & Capabilities

Architecture Type: Mixture of Experts (MoE)

Sparse Mixture-of-Experts routing: activates only a subset of parameter experts per token, delivering frontier-level intelligence with exceptional efficiency.