DeepSeek-V4-Provs.Qwen 3.8 27B

Which one is better?

Level in the overall index: the gap is too small for a clear winner. Qwen 3.8 27B is cheaper for input, DeepSeek-V4-Pro for output.

IntelligenceOverall index, 0 to 100 · higher is better

DeepSeek-V4-Pro

Rank 42 to 50

Qwen 3.8 27B

Rank 44 to 57

SpeedTokens per second · higher is better

DeepSeek-V4-Pro

fairly slow

Qwen 3.8 27B

slow

PriceDollars per 1M tokens · lower is better

$0.99

DeepSeek-V4-Pro

fairly cheap

$0.66

Qwen 3.8 27B

fairly cheap

DeepSeek-V4-Pro

Subscription: No DeepSeek subscription includes this model. Through other providers from ≈ €1.06 a month (Command Code Go, with limited credits). 20 plans in Germany.
Via API: $0.66 input, $1.98 output per 1M tokens according to the provider; the price index has listed it at $1.32 and $3.96 since Aug 17, 2026.

Qwen 3.8 27B

Subscription: No Alibaba subscription includes this model. Through other providers from ≈ €1.06 a month (Command Code Go, with limited credits). 10 plans in Germany.
Via API: $0.04 input, $2.50 output per 1M tokens according to the provider; the price index has listed it at $0.42 and $3 since Oct 27, 2026.

Overall index with data through Oct 29, 2026 · Prices from Oct 29, 2026 · Speed: measured by Artificial Analysis · Subscriptions as of Oct 24, 2026 · Subscription prices incl. VAT

Where is the difference?

Values from 0 to 100 per task area. We only name a winner when the gap is large enough and there are enough tests.

DeepSeek-V4-Pro ahead: 3Qwen 3.8 27B ahead: 2level: 3too little data: 1of 9 areas

AreaDeepSeek-V4-ProQwen 3.8 27B0 to 100, higher is betterResult

Areas in the overall index

CodingDeepSeek-V4-Pro ahead2413

Reasoninglevel166

MathematicsDeepSeek-V4-Pro ahead168

KnowledgeDeepSeek-V4-Pro ahead205

Agentslevel3421

Multimodalonly Qwen 3.8 27B has a scorenot enough data13

Specialist fields

FinanceQwen 3.8 27B ahead1934

LegalQwen 3.8 27B ahead1123

Healthlevel83

Show 2 more areas that do not count toward the overall index

AreaDeepSeek-V4-ProQwen 3.8 27B0 to 100, higher is betterResult

Cybersecurityonly Qwen 3.8 27B has a scorenot enough data19

Productivityonly DeepSeek-V4-Pro has a score27not enough data

See for yourself: same task, one attempt

SVG short test

One SVG task with the same fixed settings for every model, without hidden thinking.

Shared test round: 43b8c5c46830

DeepSeek-V4-Pro

Qwen 3.8 27B

Earlier tests

Pelican on a bicycle

A community classic for free-form SVG drawing.

DeepSeek-V4-Pro

The response reached the token limit. No complete recording is available.

Reasoning (requested): high

Current attempt

Run date: Oct 5, 2026
API provider: openrouter
Requested API model: deepseek/deepseek-v4-pro-0813
API model reported in response: deepseek/deepseek-v4-pro-0813
Reasoning setting: high
Pinned endpoint: DeepSeek | deepseek/deepseek-v4-pro-20260813
Temperature: Provider default
Token limit including reasoning: 8,192 tokens
Test protocol: community-visual-tests-v1
Technical checks: Failed
Response time: 67.6 s

Not yet visually reviewed. A successful recording does not confirm correct physics or full compliance with the prompt.

Earlier attempts (3)

Oct 2, 2026API request failedReasoning setting: high
Oct 1, 2026API request failedReasoning setting: high
Sep 7, 2026API request failedReasoning setting: high

Qwen 3.8 27B

Not tested yet

Prompt and test conditions

Generate an SVG of a pelican riding a bicycle

Token limit including reasoning: 8,192 tokens

Task origin (Simon Willison)

All data side by side

The better value is bold. For the overall index only when the gap is clear.

Metric	DeepSeek-V4-Pro	Qwen 3.8 27B
Intelligence
Overall index0 to 100, higher is better	18	11
Rank	Rank 42 to 50 of 95	Rank 44 to 57 of 95
Price per 1M tokens
Inputwhat you send	$0.66	$0.04
Outputwhat the model writes	$1.98	$2.50
Speed
Tokens per second	81	40
Technical
Context windowhow much text the model reads at once	1 M tokens≈ 1,500 A4 pages	262,144 tokens≈ 390 A4 pages
Knowledge cutoff	n/a	n/a
Released	Apr 24, 2026	Aug 14, 2026
Availability	Active	Active
Open weights	yes	yes

DeepSeek-V4-Pro: At peak times (Mon, Tue, Wed, Thu, Fri, 01:00-04:00 UTC; Mon, Tue, Wed, Thu, Fri, 06:00-10:00 UTC) the price is $1.32 input and $3.96 output per 1M tokens. The table shows the off-peak price.

Prices without caching or batch discounts · Page count as a rule of thumb from Artificial Analysis

Shared benchmarks

Single results from the same benchmark version, task and metric. A value is bold only with the same test setup and a clear gap.

A: DeepSeek-V4-ProB: Qwen 3.8 27B

AA-Omniscience AccuracyOverall · higher is better38.6%9.1%Artificial AnalysisAs of 10/29/2026

AA-Omniscience Hallucination RateOverall · lower is better93.9%18.0%Artificial AnalysisAs of 10/29/2026

Code MigrationOverall · higher is betterTest setup not documented as identical26.2%14.2%Vals AIAs of 10/18/2026

CritPt v2Overall · higher is better0.0%0.3%Artificial AnalysisAs of 10/29/2026

EMBOverall · higher is betterTest setup not documented as identical51.6%59.7%Vals AIAs of 10/20/2026

Finance Agent (v2)Overall · higher is betterTest setup not documented as identical44.1%48.6%Vals AIAs of 10/20/2026

GDPval-AAOverall · higher is better1,066.48 elo1,082.47 eloArtificial AnalysisAs of 10/29/2026

GPQA DiamondOverall · higher is betterTest setup not documented as identical89.4%88.9%Vals AIAs of 09/01/2026

Legal Research BenchOverall · All-pass · higher is betterTest setup not documented as identical23.1%36.1%Vals AIAs of 10/26/2026

LegalBenchOverall · higher is betterTest setup not documented as identical80.3%82.4%Vals AIAs of 10/13/2026

Show 54 more benchmarks

LiveCodeBenchOverall · higher is betterTest setup not documented as identical87.5%84%Vals AIAs of 09/01/2026

MedCodeOverall · higher is betterTest setup not documented as identical40.5%28.7%Vals AIAs of 10/13/2026

MedScribeOverall · higher is betterTest setup not documented as identical75.1%83.8%Vals AIAs of 10/13/2026

MMLU ProOverall · higher is betterTest setup not documented as identical87.2%84.3%Vals AIAs of 09/01/2026

ProgramBenchOverall · higher is betterTest setup not documented as identical0%0%Vals AIAs of 10/20/2026

ProofBench v1.1Overall · higher is betterTest setup not documented as identical16%16%Vals AIAs of 10/24/2026

SkillsBenchOverall · higher is betterTest setup not documented as identical51.3%38.1%Vals AIAs of 09/27/2026

SWE-bench VerifiedOverall · higher is betterTest setup not documented as identical77.4%86%Vals AIAs of 09/01/2026

Tax Agent BenchOverall · higher is betterTest setup not documented as identical25.2%30.9%Vals AIAs of 10/26/2026

TaxEval v2Overall · higher is betterTest setup not documented as identical72.1%70.9%Vals AIAs of 09/01/2026

Terminal-Bench 2.1Overall · higher is betterTest setup not documented as identical50.2%58.4%Vals AIAs of 10/01/2026

Vibe Code Bench v1.1Overall · higher is betterTest setup not documented as identical49.9%64.9%Vals AIAs of 10/26/2026

Code MigrationCLI · higher is betterTest setup not documented as identical15.5%2.2%Vals AIAs of 10/18/2026

EMBComps · higher is betterTest setup not documented as identical57.4%53.3%Vals AIAs of 10/20/2026

Finance Agent (v2)Adjustments · higher is betterTest setup not documented as identical41.6%47.3%Vals AIAs of 10/20/2026

GPQA DiamondFew-Shot CoT · higher is betterTest setup not documented as identical90.4%88.9%Vals AIAs of 09/01/2026

Legal Research BenchAdministrative / Regulatory · All-pass · higher is betterTest setup not documented as identical28.8%45.5%Vals AIAs of 10/26/2026

LegalBenchConclusion Tasks · higher is betterTest setup not documented as identical89.1%88.4%Vals AIAs of 10/13/2026

LiveCodeBenchEasy · higher is betterTest setup not documented as identical98.8%99.1%Vals AIAs of 09/01/2026

MMLU ProBiology · higher is betterTest setup not documented as identical92.9%91.5%Vals AIAs of 09/01/2026

ProgramBenchAlmost Resolved · higher is betterTest setup not documented as identical0%1%Vals AIAs of 10/20/2026

SkillsBenchNo Skills · higher is betterTest setup not documented as identical27.2%19.1%Vals AIAs of 09/27/2026

SWE-bench Verified<15 min fix · higher is betterTest setup not documented as identical87.6%90.2%Vals AIAs of 09/01/2026

Tax Agent BenchControversy & Precedence Analysis · higher is betterTest setup not documented as identical12.3%6.7%Vals AIAs of 10/26/2026

TaxEval v2Correctness · higher is betterTest setup not documented as identical60.6%60.1%Vals AIAs of 09/01/2026

Terminal-Bench 2.1Easy · higher is betterTest setup not documented as identical66.7%50%Vals AIAs of 10/01/2026

Code MigrationCOBOL · higher is betterTest setup not documented as identical58.4%50%Vals AIAs of 10/18/2026

EMBDataroom · higher is betterTest setup not documented as identical62.6%79.7%Vals AIAs of 10/20/2026

Finance Agent (v2)All-Pass · higher is betterTest setup not documented as identical31.4%36.8%Vals AIAs of 10/20/2026

GPQA DiamondZero-Shot CoT · higher is betterTest setup not documented as identical88.4%88.9%Vals AIAs of 09/01/2026

Legal Research BenchAdministrative / Regulatory · Weighted partial credit · higher is betterTest setup not documented as identical83.4%88.5%Vals AIAs of 10/26/2026

LegalBenchInterpretation Tasks · higher is betterTest setup not documented as identical75.2%82%Vals AIAs of 10/13/2026

LiveCodeBenchHard · higher is betterTest setup not documented as identical72.6%65.7%Vals AIAs of 09/01/2026

MMLU ProBusiness · higher is betterTest setup not documented as identical91.9%90.2%Vals AIAs of 09/01/2026

ProgramBenchFully Resolved · higher is betterTest setup not documented as identical0%0%Vals AIAs of 10/20/2026

SkillsBenchWith Skills · higher is betterTest setup not documented as identical51.3%38.1%Vals AIAs of 09/27/2026

SWE-bench Verified>4 hours · higher is betterTest setup not documented as identical0%66.7%Vals AIAs of 09/01/2026

Tax Agent BenchControversy & Precedence Analysis · higher is betterTest setup not documented as identical54.8%48.3%Vals AIAs of 10/26/2026

TaxEval v2Stepwise Reasoning · higher is betterTest setup not documented as identical83.6%81.6%Vals AIAs of 09/01/2026

Terminal-Bench 2.1Hard · higher is betterTest setup not documented as identical31.1%43.3%Vals AIAs of 10/01/2026

Code MigrationCode Quality · higher is betterTest setup not documented as identical3.401 / 54.197 / 5Vals AIAs of 10/18/2026

EMBDCF · higher is betterTest setup not documented as identical44.4%50%Vals AIAs of 10/20/2026

Finance Agent (v2)Comparables · higher is betterTest setup not documented as identical39.6%38.9%Vals AIAs of 10/20/2026

Legal Research BenchBusiness & Commercial · All-pass · higher is betterTest setup not documented as identical23.3%41.9%Vals AIAs of 10/26/2026

LegalBenchIssue Tasks · higher is betterTest setup not documented as identical84.6%90.2%Vals AIAs of 10/13/2026

LiveCodeBenchMedium · higher is betterTest setup not documented as identical91.1%87.2%Vals AIAs of 09/01/2026

MMLU ProChemistry · higher is betterTest setup not documented as identical92.5%91.8%Vals AIAs of 09/01/2026

ProgramBenchRaw Pass Rate · higher is betterTest setup not documented as identical47.8%11.2%Vals AIAs of 10/20/2026

SWE-bench Verified1-4 hours · higher is betterTest setup not documented as identical45.2%69%Vals AIAs of 09/01/2026

Tax Agent BenchCurrent & Temporal Analysis · higher is betterTest setup not documented as identical36.6%43.1%Vals AIAs of 10/26/2026

Terminal-Bench 2.1Medium · higher is betterTest setup not documented as identical59.4%67.3%Vals AIAs of 10/01/2026

EMBLBO · higher is betterTest setup not documented as identical43.9%53.7%Vals AIAs of 10/20/2026

Finance Agent (v2)Disclosure Analysis · higher is betterTest setup not documented as identical52.6%60.3%Vals AIAs of 10/20/2026

Legal Research BenchBusiness & Commercial · Weighted partial credit · higher is betterTest setup not documented as identical86.5%86.8%Vals AIAs of 10/26/2026

Compare another matchup

First modelSecond model

More matchups

Compare DeepSeek-V4-Pro and Qwen 3.8 27B with other leading models.

View the complete LLM comparison

DeepSeek-V4-Provs.Qwen 3.8 27B

DeepSeek-V4-Pro

Qwen 3.8 27B

Where is the difference?

See for yourself: same task, one attempt

SVG short test

DeepSeek-V4-Pro

Qwen 3.8 27B

Pelican on a bicycle

All data side by side

Shared benchmarks

Compare another matchup

Related LLM comparisons