DeepSeek V4 Pro vs GPT-OSS-120B vs Qwen3.8-Flash-Next vs Hy4 flash

New Comparison

Scores are fetched live from the current database on every visit — nothing is cached.

Winner

DeepSeek V4 Pro

DeepSeek V4 Pro scores higher overall across category strength profiles (81 vs 69 for Qwen3.8-Flash-Next); only 0 benchmarks are directly comparable.

DeepSeek V4 Pro 80.5 Qwen3.8-Flash-Next 69.1 Hy4 flash 62.9 GPT-OSS-120B 40.1

Only 0 of 111 benchmarks are scored by every selected model — the verdict leans on category strength profiles and is indicative.

Benchmark Scores Comparison

DeepSeek V4 Pro GPT-OSS-120B Qwen3.8-Flash-Next Hy4 flash

DeepSeek V4 Pro

Score: 15.0

GPT-OSS-120B

116.8B params

Score: 100.0

Qwen3.8-Flash-Next

125.0B params

Score: 70.0

Hy4 flash

220.0B params

Coverage matrix

4 models × 111 benchmarks — hover a benchmark name for its category

Full Coverage Matrix
Hides benchmarks where all selected models score identically.
Shading = score relative to the best score among the selected models: Low Mid High Missing (—) The totals row shows how many of the 111 benchmarks each model has scores for.
Benchmark GPT-OSS-120B OpenAI Qwen3.8-Flash-Next Qwen Team (Alibaba) Hy4 flash Tencent DeepSeek V4 Pro DeepSeek
AGIEval (Chinese) — — 90.6 —
SVG-Bench — — — 61.7
GPQA Diamond 80.1 91.7 — 92.8
CountBenchQA — — 86.6 —
OCW Courses — — 64 —
ArXivMath — — — 62.1
Chinese-SimpleQA — — 74.6 —
LiveCodeBench v6 — 91.9 — —
Hy-LifeSearch (Internal) — — — 46.9
Aider Polyglot 44.4 — — —
Automation-Bench — — — 30.4
SWE-bench Pro — 62.5 — 55.4
RealWorldQA — 88.5 71.1 —
Agents' Last Exam (Score) — 51.2 — —
BioMysteryBench — — — 61.6
AA-LCR — 79.1 — —
MathVision — 90.6 — —
HealthBench Hard 30 — — —
Tau-Bench Retail 67.8 — — —
AndroidWorld — 84.5 — —
Agents' Last Exam — 24.3 — 21.9
Toolathlon Verified — 73.5 — 74.1
Hy-Backend 2.0 (Internal) — — — 33.8
RecreationBench — 49.9 — —
Hy-SWE Max Verified (Internal) — — — 65.7
Hy-BrowseComp-Pro2 (Internal) — — — 46.5
ARC-c — — 95 —
GDPVal — — — 70.3
OfficeQA Pro — — — 65.4
PIQA — — 84.4 —
SWE Atlas - TW — — — 45.6
MBPP+ — — 73.7 —
WorkBuddy Blind User Evaluation (satisfaction) — — 2.79 —
SWE Atlas - QnA — — — 53.4
TextVQA — — 73.4 —
NL2Repo — — — 61.5
MMMU-Pro — — 52 —
PostTrainBench — — — 24.5
INCLUDE — — 82.8 —
MMMLU 81.3 — — —
Claw-Eval Avg — 60.4 — 58.4
Apex-Agents — — — 32.4
HealthBench Consensus 89.9 — — —
WorkSpaceBench — — — 65.4
MMLU-Pro — — 71.1 —
CodeForces 2463 — — —
BankerToolBench — — — 73.1
SuperGPQA — — 57.8 —
E-Bench-Code (Internal) — — — 64.3
CV-Bench — — 83.4 —
HellaSwag — — 85.2 —
HLE (with tools) 19 35.9 — 60
LVBench — 76.6 — —
CharXiv (RQ) — 84.6 — —
WorkIndex (Tencent internal) — — 42.1 —
AIME 2024 95.8 — — —
Belebele (5-language subset) — — 93.8 —
KOR-Bench — — 54.7 —
CMMLU — — 91.6 —
CritPt (no tools) — — — 18
BrowseComp — — — 83.4
MMLU-Redux — — 87.6 —
NL2Repo-Bench — 48.1 — 35.5
DRACO — — — 77.3
CoWorkBench — 73.9 — —
SWE-bench Multilingual — 81 — 77.3
MMLU 90 — 88.7 —
YC-Bench — — — 500000
Hy-FinAgentBench (Internal) — — — 78.5
IFBench — 81.3 — —
SWE-bench Verified 62.4 — — 80.6
Humanity's Last Exam 14.9 — — 42.7
ERQA — 72.3 — —
U-MATH — — 69.8 —
SpreadSheetBench-v1 — — — 84.9
SUPERChem — — — 62
VIBE-V2 — — — 45
HorizonMath (pass@4) — — — 4.42
Terminal-Bench 2.1 (Terminus-2) — 83.3 — 87.9
DeepSWE — — — 62.7
BrokenArXiv — — — 43.1
SocialIQA — — 80.1 —
Hy-FinmodelBench v2 (Internal) — — — 51.3
MathArena Apex 2025 — — — 66.3
Terminal Bench 2.1 — — — 59.6
AIME 2025 92.5 — — —
JobBench — 55.7 — 54.1
$OneMillion-Bench — — — 62
OSWorld 2.0 (Partial) — 52.3 — —
OSWorld 2.0 (Binary) — 19.4 — —
C-Eval — — 92 —
WideSearch — — — 81.8
Tau-Bench Airline 49.2 — — —
SimpleVQA — — 48.5 —
SkillsBench Avg5 — — — 65
Vision2Web — 64 — —
HealthBench 57.6 — — —
Claw-Eval Pass^3 — 64.4 — —
E-Bench (Internal) — — — 61.3
ProgramBench — — — 15.5
MCP-Atlas — — — 73.6
SWE-Marathon — — — 19
Cybergym — — — 83.3
SWE Atlas - RF — — — 48.6
DeepSWE 1.1 — 58.7 — —
Hy-CompanyBench V2 (Internal) — — — 64.4
C³ — — 77.3 —
Harbor-Index — — — 36.9
AGIEval (English) — — 75.4 —
GDPVal-AA v2 — — — 1580
SimpleQA — — 37.5 —
Covered 15/111 27/111 29/111 56/111

Scorecard

Overall 0–100 score — category strengths (50%), shared-benchmark wins (30%), coverage (20%)

Model Overall score Shared benchmark wins Coverage Strongest at Weakest at
DeepSeek V4 Pro Winner 80.5 — 56 of 111 (50.0%) STEM & Reasoning Agentic Coding
Qwen3.8-Flash-Next 69.1 — 27 of 111 (24.0%) Vision & Language STEM & Reasoning
Hy4 flash 62.9 — 29 of 111 (26.0%) Knowledge Vision & Language
GPT-OSS-120B 40.1 — 15 of 111 (14.0%) Knowledge STEM & Reasoning

Category strengths

Where each model stands, per benchmark category — score relative to the best model, adjusted for benchmark coverage. This also ranks models when they share few benchmarks.

Category DeepSeek V4 Pro Qwen3.8-Flash-Next Hy4 flash GPT-OSS-120B
General Agents 30 benchmarks 92.6 99.6 — —
Agentic Coding 25 benchmarks 88.8 91.2 — 48.3
STEM & Reasoning 16 benchmarks 100 61.9 — 39.3
Knowledge 8 benchmarks — — 95.7 100
Vision & Language 6 benchmarks — 100 59.5 —

Partial coverage benchmarks

Scored by at least one but not all selected models (111) — missing scores show as —

Benchmark Category DeepSeek V4 Pro GPT-OSS-120B Qwen3.8-Flash-Next Hy4 flash
AGIEval (Chinese) 1/4 models reasoning — — — 90.6
SVG-Bench 1/4 models coding_agent 61.7 — — —
GPQA Diamond 3/4 models stem_reasoning 92.8 80.1 91.7 —
CountBenchQA 1/4 models multimodal — — — 86.6
OCW Courses 1/4 models stem_reasoning — — — 64
ArXivMath 1/4 models stem_reasoning 62.1 — — —
Chinese-SimpleQA 1/4 models knowledge — — — 74.6
LiveCodeBench v6 1/4 models stem_reasoning — — 91.9 —
Hy-LifeSearch (Internal) 1/4 models general_agent 46.9 — — —
Aider Polyglot 1/4 models coding_agent — 44.4 — —
Automation-Bench 1/4 models general_agent 30.4 — — —
SWE-bench Pro 2/4 models coding_agent 55.4 — 62.5 —
RealWorldQA 2/4 models vision_language — — 88.5 71.1
Agents' Last Exam (Score) 1/4 models general_agent — — 51.2 —
BioMysteryBench 1/4 models stem_reasoning 61.6 — — —
AA-LCR 1/4 models long_context — — 79.1 —
MathVision 1/4 models vision_language — — 90.6 —
HealthBench Hard 1/4 models general_capabilities — 30 — —
Tau-Bench Retail 1/4 models general_agent — 67.8 — —
AndroidWorld 1/4 models general_agent — — 84.5 —
Agents' Last Exam 2/4 models general_agent 21.9 — 24.3 —
Toolathlon Verified 2/4 models general_agent 74.1 — 73.5 —
Hy-Backend 2.0 (Internal) 1/4 models coding_agent 33.8 — — —
RecreationBench 1/4 models general_agent — — 49.9 —
Hy-SWE Max Verified (Internal) 1/4 models coding_agent 65.7 — — —
Hy-BrowseComp-Pro2 (Internal) 1/4 models general_agent 46.5 — — —
ARC-c 1/4 models reasoning — — — 95
GDPVal 1/4 models general_agent 70.3 — — —
OfficeQA Pro 1/4 models general_agent 65.4 — — —
PIQA 1/4 models reasoning — — — 84.4
SWE Atlas - TW 1/4 models coding_agent 45.6 — — —
MBPP+ 1/4 models code — — — 73.7
WorkBuddy Blind User Evaluation (satisfaction) 1/4 models general_capabilities — — — 2.79
SWE Atlas - QnA 1/4 models coding_agent 53.4 — — —
TextVQA 1/4 models multimodal — — — 73.4
NL2Repo 1/4 models coding_agent 61.5 — — —
MMMU-Pro 1/4 models vision_language — — — 52
PostTrainBench 1/4 models coding_agent 24.5 — — —
INCLUDE 1/4 models multilingual — — — 82.8
MMMLU 1/4 models multilingual — 81.3 — —
Claw-Eval Avg 2/4 models coding_agent 58.4 — 60.4 —
Apex-Agents 1/4 models general_agent 32.4 — — —
HealthBench Consensus 1/4 models general_capabilities — 89.9 — —
WorkSpaceBench 1/4 models general_agent 65.4 — — —
MMLU-Pro 1/4 models knowledge — — — 71.1
CodeForces 1/4 models stem_reasoning — 2463 — —
BankerToolBench 1/4 models general_agent 73.1 — — —
SuperGPQA 1/4 models knowledge — — — 57.8
E-Bench-Code (Internal) 1/4 models coding_agent 64.3 — — —
CV-Bench 1/4 models vision_language — — — 83.4
HellaSwag 1/4 models reasoning — — — 85.2
HLE (with tools) 3/4 models stem_reasoning 60 19 35.9 —
LVBench 1/4 models video_understanding — — 76.6 —
CharXiv (RQ) 1/4 models document_understanding — — 84.6 —
WorkIndex (Tencent internal) 1/4 models general_agent — — — 42.1
AIME 2024 1/4 models stem_reasoning — 95.8 — —
Belebele (5-language subset) 1/4 models multilingual — — — 93.8
KOR-Bench 1/4 models reasoning — — — 54.7
CMMLU 1/4 models knowledge — — — 91.6
CritPt (no tools) 1/4 models stem_reasoning 18 — — —
BrowseComp 1/4 models general_agent 83.4 — — —
MMLU-Redux 1/4 models knowledge — — — 87.6
NL2Repo-Bench 2/4 models coding_agent 35.5 — 48.1 —
DRACO 1/4 models general_agent 77.3 — — —
CoWorkBench 1/4 models general_agent — — 73.9 —
SWE-bench Multilingual 2/4 models coding_agent 77.3 — 81 —
MMLU 2/4 models knowledge — 90 — 88.7
YC-Bench 1/4 models general_agent 500000 — — —
Hy-FinAgentBench (Internal) 1/4 models domain_finance 78.5 — — —
IFBench 1/4 models instruction_following — — 81.3 —
SWE-bench Verified 2/4 models coding_agent 80.6 62.4 — —
Humanity's Last Exam 2/4 models stem_reasoning 42.7 14.9 — —
ERQA 1/4 models spatial_intelligence — — 72.3 —
U-MATH 1/4 models stem_reasoning — — — 69.8
SpreadSheetBench-v1 1/4 models general_agent 84.9 — — —
SUPERChem 1/4 models stem_reasoning 62 — — —
VIBE-V2 1/4 models coding_agent 45 — — —
HorizonMath (pass@4) 1/4 models stem_reasoning 4.42 — — —
Terminal-Bench 2.1 (Terminus-2) 2/4 models coding_agent 87.9 — 83.3 —
DeepSWE 1/4 models coding_agent 62.7 — — —
BrokenArXiv 1/4 models stem_reasoning 43.1 — — —
SocialIQA 1/4 models reasoning — — — 80.1
Hy-FinmodelBench v2 (Internal) 1/4 models domain_finance 51.3 — — —
MathArena Apex 2025 1/4 models stem_reasoning 66.3 — — —
Terminal Bench 2.1 1/4 models coding_agent 59.6 — — —
AIME 2025 1/4 models stem_reasoning — 92.5 — —
JobBench 2/4 models general_agent 54.1 — 55.7 —
$OneMillion-Bench 1/4 models general_capabilities 62 — — —
OSWorld 2.0 (Partial) 1/4 models general_agent — — 52.3 —
OSWorld 2.0 (Binary) 1/4 models general_agent — — 19.4 —
C-Eval 1/4 models knowledge — — — 92
WideSearch 1/4 models general_agent 81.8 — — —
Tau-Bench Airline 1/4 models general_agent — 49.2 — —
SimpleVQA 1/4 models vision_language — — — 48.5
SkillsBench Avg5 1/4 models coding_agent 65 — — —
Vision2Web 1/4 models vision_language — — 64 —
HealthBench 1/4 models general_capabilities — 57.6 — —
Claw-Eval Pass^3 1/4 models coding_agent — — 64.4 —
E-Bench (Internal) 1/4 models general_agent 61.3 — — —
ProgramBench 1/4 models coding_agent 15.5 — — —
MCP-Atlas 1/4 models general_agent 73.6 — — —
SWE-Marathon 1/4 models coding_agent 19 — — —
Cybergym 1/4 models general_agent 83.3 — — —
SWE Atlas - RF 1/4 models coding_agent 48.6 — — —
DeepSWE 1.1 1/4 models coding_agent — — 58.7 —
Hy-CompanyBench V2 (Internal) 1/4 models general_agent 64.4 — — —
C³ 1/4 models reasoning — — — 77.3
Harbor-Index 1/4 models coding_agent 36.9 — — —
AGIEval (English) 1/4 models reasoning — — — 75.4
GDPVal-AA v2 1/4 models general_agent 1580 — — —
SimpleQA 1/4 models knowledge — — — 37.5

Comparison FAQ

Which model scores highest in this comparison?

DeepSeek V4 Pro scores higher overall across category strength profiles (81 vs 69 for Qwen3.8-Flash-Next); only 0 benchmarks are directly comparable.

Which model is strongest at Agentic Coding?

Qwen3.8-Flash-Next leads the Agentic Coding category (strength score 91.2/100 across 6 comparable benchmarks).

How reliable is this comparison verdict?

Only 0 of 111 benchmarks are scored by every selected model — the verdict leans on category strength profiles and is indicative.

What now?

You've seen 4 models side by side — keep the loop going.

Share this comparison

The URL is the comparison — copy or share it anywhere. No tracking, no cookies.

Not the right matchup?

Open the create form with these models pre-filled and untick the one you want to replace.

Swap a model