Atlas

Compare Models

DeepSeek V4.1 Flash 1017 vs GPT-5.6 Sol

GPT-5.6 Sol leads by 1.9 capability points (≈95% interval −0.4 to +4.3): statistically tied · P(higher) ≈ 94%

Wins 70 of 106 shared benchmarks (35 lost, 1 tied)

Capability vs. relative cost
DeepSeek V4.1 Flash 1017GPT-5.6 SolClosed weightsOpen weights

Each line runs through one model’s priced reasoning settings, cheapest to most expensive; grey points are other models’ priced settings in the same range. Results whose reasoning effort the source didn’t report are left out wherever a model has named effort levels. Capability and relative cost are the published estimates (relative cost is cost per task as a multiple of GPT-5 (high)); hover a setting, or use the arrow keys, to set it beside the other model’s same setting.

DeepSeek V4.1 Flash 1017vsGPT-5.6 Solcapability vs. relative costData Nov 15, 2026Village LLM Atlasatlas.village.dev

Benchmark wins35 – 70 · 106 shared, 1 tied

Comparing matching setups only (36 benchmarks excluded) · suites counted once
← DeepSeek V4.1 Flash 1017GPT-5.6 Sol →

DeepSeek V4.1 Flash 1017vsGPT-5.6 Solwins by categoryData Nov 15, 2026Data: Vals AI, Artificial Analysis, SecureBio and 30 othersVillage LLM Atlasatlas.village.dev

Margins

Each benchmark scaled 0–100% (its declared range, else the field’s range); Δ is GPT-5.6 Sol minus DeepSeek V4.1 Flash 1017.

LisanBench Difficulty-WeightedGames77331403−67.8
HardGeoBench Avg DistanceMultimodal32011429+54.7
LisanBench Path LengthGames211074699−51.0
RiemannBench (November 2026 rescoring)Math38.483.2+44.8
FrontierMath Tier 4 (v2) - PrivateMath46.382.9+36.6
HardGeoBench <=100miMultimodal7.541.3+33.8
HardGeoBench Country AccuracyMultimodal15.047.5+32.5
CritPt v2 (agentic, avg@5)Science50.677.1+26.6
HardGeoBench <=10miMultimodal3.830.0+26.3
Concept Synth INDUCTION Challenge64 Round 1General QA32.857.8+25.0
Concept Synth INDUCTION Challenge100 Round 1General QA21.043.0+22.0
FrontierMath Tiers 1-3 (v2) - PrivateMath71.289.1+17.9
BuseyBench SVGMultimodal4.36.1+17.6
HPCT (SecureBio)Science60.576.3+15.8
NoCoT-Bench NCRI 15.2General QA81.6109+15.3
WCB (SecureBio)Science49.664.4+14.8
AutomationBench 1.0.6 (provider-reported)Agents60.545.8−14.7
Bio Capabilities Index (SecureBio)Indexes149169+14.5
ARC-AGI-2General QA78.192.5+14.4
ALE-BenchCode20122410+14.1
GBENCH One-shot Coding GScoreGames53.867.7+13.9
NoCoT-Bench NCKI (kspine v3)General QA102121+13.2
Chess Puzzles (task version 2.0)Games46.059.0+13.0
ReactBench (39 tasks)Code34.446.7+12.3
EnterpriseBench: CoreCraftProfessional Work58.546.2−12.3
PerceptionBenchMultimodal47.759.7+12.0
NL2Repo - release-reported repository generation scoreCode68.256.8−11.4
VCT (SecureBio)Science42.954.3+11.4
τ³-Banking (Fireworks, banking_knowledge)Agents46.835.7−11.2
VCT v2 (SecureBio)Science38.549.4+10.9
GDP.pdf (Artificial Analysis, All-pass)Professional Work17.227.8+10.6
Excel Modeling BenchmarkFinance & Accounting61.872.3+10.5
ZeroBench Main Mean pass@1 (five samples)Multimodal11.822.0+10.2
GameBench 2Games69.980.1+10.2
Humanity's Last Exam DiamondGeneral QA20.830.4+9.6
Terminal-Bench-Science 0.1Science12.922.4+9.5
Finance Agent v2 - General Qualitative AnalysisFinance & Accounting74.164.9−9.2
Finance Agent v2 - AdjustmentsFinance & Accounting58.449.2−9.1
Mystery Game PuzzlesGames49.058.0+9.0
Code MigrationCode44.252.9+8.7
Agents' Last Exam - release-reported pass rate, subset unspecifiedAgents35.426.7−8.7
AA-Omniscience IndexIndexes4.722.0+8.6
AutomationBench-AA 1.0.6Agents68.560.1−8.4
CharXiv ReasoningMultimodal77.084.6+7.6
ARC-AGI-3 (Semi-Private)General QA0.27.8+7.6
Vibe Code Bench v1.1Code87.780.5−7.2
MBCT (SecureBio)Science53.961.1+7.2
Tax Agent BenchFinance & Accounting74.968.0−7.0
Harvey's Legal Agent BenchmarkLegal9.22.5−6.7
DeepSWE v1.1Code77.070.7−6.3
ProgramBench - almost solved rollout percentageCode29.123.0−6.1
Terminal-Bench 4.0Code31.837.9+6.1
Terminal-Bench 4.0 (Artificial Analysis)Code33.839.9+6.1
CursorBench 4.0Code47.741.7−6.0
WeirdML v3Science9.215.0+5.8
Bedside BenchHealth & Medicine85.791.1+5.4
RedlineBenchLegal44.950.3+5.4
DeepsecBench — Revalidated V1 (233 issues)Code30.035.3+5.3
LMArena Agent Arena Net Improvement (Steering Burden)Agents2.04.6+5.0
CritPt (avg@5)Science24.028.6+4.6
LMArena Agent Arena Bash ReliabilityAgents2.411.4+4.5
LAB-Bench CloningScenarios (SecureBio)Science94.898.8+4.0
Wiz Cyber Model Arena v3.9Cybersecurity60.364.2+3.9
Finance Agent v2 - All PassFinance & Accounting45.241.4−3.9
LAB-Bench LitQA2 (SecureBio)Science49.052.8+3.8
BabyVision with PythonMultimodal92.588.9−3.6
BabyVision with toolsMultimodal92.588.9−3.6
Finance Agent v2 - General Quantitative AnalysisFinance & Accounting79.275.6−3.6
Medical Long Context Reasoning (MLCR-AA, October 2026 methodology)Health & Medicine26.730.0+3.3
Ruby LLM Benchmarks — Program FixerCode72.075.2+3.2
Chartography (with tools)Multimodal83.079.9−3.1
Political Science LLM BenchmarkProfessional Work66.269.2+3.0
ARC-AGI-1 Semi-PrivateGeneral QA94.597.5+3.0
Epoch Capabilities Index (ECI)Indexes159163+3.0
MMMU ProMultimodal79.782.7+2.9
LiveBench 2.0Agents79.276.3−2.9
Finance Agent v2 - Disclosure AnalysisFinance & Accounting66.363.3−2.9
Humanity's Last ExamGeneral QA42.645.5+2.9
Legal Research BenchLegal45.248.1+2.9
Artificial Analysis Progress Index v1Indexes324331+2.7
LMArena Agent Arena Praise vs ComplaintAgents1.26.3+2.6
AA-Briefcase v1.1 EloAgents14161369−2.6
GPQA-Bio (SecureBio)Science87.885.3−2.5
Finance Agent v2 - ComparablesFinance & Accounting47.545.3−2.3
Finance Agent v2 - Financial ModelingFinance & Accounting22.820.7−2.0
Artificial Analysis Intelligence Index v4.3.2Indexes45.147.0+1.9
SciCode 1.0.1 (Artificial Analysis)Code55.957.8+1.9
WMDP-Bio (SecureBio)Science87.989.7+1.8
DAYJOB: Healthcare (October 2026 rescoring)Health & Medicine6.04.4−1.6
Finance Agent v2 - Market AnalysisFinance & Accounting62.764.2+1.5
Vals Index v2.1Indexes56.658.0+1.4
AA-LCR v1.1Search82.784.0+1.3
Bug Hunt BenchCode40.241.4+1.2
APEX-SWECode49.148.0−1.1
PubMedQA (SecureBio)Health & Medicine80.081.1+1.1
Finance Agent v2 - Earnings AnalysisFinance & Accounting75.374.4−0.9
LMArena Agent Arena Steering BurdenAgents4.05.5+0.7
DAYJOB: Finance (October 2026 rescoring)Finance & Accounting12.511.8−0.7
OTIS Mock AIME 2024-2025Math99.4100.0+0.6
MMLU-Bio (SecureBio)Science90.791.1+0.4
LMArena Agent Arena Tool HallucinationAgents0.31.0+0.4
Finance Agent v2 - PrecedentsFinance & Accounting26.526.2−0.3
Humanity's Sixth SenseMultimodal30.130.0−0.1
FoodTruck BenchGames5328953229−0.1
LMArena Agent Arena Confirmed SuccessAgents2.32.30.0
LAB-Bench ProtocolQA (SecureBio)Science81.981.90.0

DeepSeek V4.1 Flash 1017vsGPT-5.6 Solnormalised score marginsData Nov 15, 2026Data: Vals AI, Artificial Analysis, SecureBio and 30 othersVillage LLM Atlasatlas.village.dev

Win counts use each benchmark’s declared score direction. Represented suite slices and the derived Capability score are excluded from headline totals. Missing or incompatible setups are excluded, never counted as losses.
Loading Atlas data…