Enter opens the highlighted result. Esc closes.

DeepSeek V3.1 Terminus vs GPT-4

Updated October 21, 2026. 27 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.

Replace DeepSeek V3.1 Terminus Replace GPT-4
DeepSeek V3.1 TerminusDeepSeek
Coding
No data
Overall
14.8
Output $/M
$1.0142
27
GPT-4OpenAI
Coding
No data
Overall
6.7
Output $/M
$60.00
0
vs boards won of 27 shared

Decision reading

Neither model has a coding score yet, so the two cannot be ranked on coding.

DeepSeek V3.1 Terminus has the edge where output tokens cost 59.2x less, and where the context window is larger, 164K against 8K tokens.

Model ADeepSeek V3.1 TerminusDeepSeekNot benchmarkedNo coding benchmarkNo cost per solved taskModel BGPT-4OpenAINot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
27
DeepSeek V3.1 Terminus only
17
GPT-4 only
18
Like-for-like categories
6 of 9

Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.

Which one for your work

Each card is decided by one number both models carry. Where only one has it, the card says so and makes no call.

At a glance

Coding index
DeepSeek V3.1 Terminus not benchmarked, GPT-4 not benchmarked.
Widest category
Knowledge, 43.5 points apart on average over 5 shared benchmarks.
Widest single gap
LMArena Text: medicine and healthcare: DeepSeek V3.1 Terminus 1468 against GPT-4 1145, 53.9 points apart on a 0 to 100 scale.
Output price
DeepSeek V3.1 Terminus $1.0142 against GPT-4 $60.00 per million output tokens, 59.2x apart.
Context window
DeepSeek V3.1 Terminus takes 164K tokens, GPT-4 8K.

Benchmark by benchmark

First the score: boards each model leads, per category, where both have a result. Then every board, a category to a fold; bars are each result on a 0 to 100 scale, so compare the two in a row, not one row with another.

Boards won, where both have a result
CategoryDeepSeek V3.1 TerminusBarsGPT-4Ties
Coding30
Intelligence10
Reasoning30
Math10
Knowledge50
General140
All shared boards270
CodingLike-for-likeSix of these boards feed the Vibe Coding Index30
Coding benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
SWE-bench VerifiedNot measured22.4%Only GPT-4 measured
SWE-bench LiteNot measured39.7%Only GPT-4 measured
SWE-bench (full test set)Not measured12.5%Only GPT-4 measured
AA Coding Index43.513.1DeepSeek V3.1 Terminus
LiveCodeBench (Artificial Analysis run)79.8%Not measuredOnly DeepSeek V3.1 Terminus measured
SciCode38.0%Not measuredOnly DeepSeek V3.1 Terminus measured
Terminal-Bench Hard (Artificial Analysis run)31.8%Not measuredOnly DeepSeek V3.1 Terminus measured
ALE-Bench745.2Not measuredOnly DeepSeek V3.1 Terminus measured
LMArena Text: coding14251209DeepSeek V3.1 Terminus
LMArena Text: software and IT14261195DeepSeek V3.1 Terminus
Neither model has a result on the other 28 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
Intelligence benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
AA Intelligence Index14.86.7DeepSeek V3.1 Terminus
Epoch Capabilities IndexNot measured125.7Only GPT-4 measured
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
Agentic work benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
AA Agentic Index6.8Not measuredOnly DeepSeek V3.1 Terminus measured
τ²-Bench Telecom37.1%Not measuredOnly DeepSeek V3.1 Terminus measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index30
Reasoning benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
GPQA Diamond (Artificial Analysis run)79.2%34.9%DeepSeek V3.1 Terminus
Humanity's Last Exam (Artificial Analysis run)16.4%Not measuredOnly DeepSeek V3.1 Terminus measured
BIG-Bench HardNot measured75.1%Only GPT-4 measured
LMArena Text: hard prompts14261200DeepSeek V3.1 Terminus
LMArena Text: expertNot measured1149Only GPT-4 measured
LMArena Text: hard prompts in English14321209DeepSeek V3.1 Terminus
Neither model has a result on the other 15 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index10
Math benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
AA Math Index89.7Not measuredOnly DeepSeek V3.1 Terminus measured
AIME 2025 (Artificial Analysis run)89.7%Not measuredOnly DeepSeek V3.1 Terminus measured
GSM8KNot measured92.0%Only GPT-4 measured
LMArena Text: math14021230DeepSeek V3.1 Terminus
LMArena Text: mathematical workNot measured1228Only GPT-4 measured
Neither model has a result on the other 8 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index50
Knowledge benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
MMLU Pro (Artificial Analysis run)85.1%56.2%DeepSeek V3.1 Terminus
MMLUNot measured86.4%Only GPT-4 measured
AI2 Reasoning ChallengeNot measured96.3%Only GPT-4 measured
LMArena Text: science14511177DeepSeek V3.1 Terminus
LMArena Text: medicine and healthcare14681145DeepSeek V3.1 Terminus
LMArena Text: legal and government14571191DeepSeek V3.1 Terminus
LMArena Text: business and finance14141145DeepSeek V3.1 Terminus
Neither model has a result on the other 10 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
Web design benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
Design Arena: websites1187Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: UI components1182Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: game dev1131Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: data viz1167Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: 3D1141Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: all code categories1175Not measuredOnly DeepSeek V3.1 Terminus measured
Design Arena: SVG1065Not measuredOnly DeepSeek V3.1 Terminus measured
Neither model has a result on the other 2 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index140
General benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
IFBench57.0%33.2%DeepSeek V3.1 Terminus
Long context reasoning (AA-LCR)69.3%Not measuredOnly DeepSeek V3.1 Terminus measured
DTBench81.3%62.7%DeepSeek V3.1 Terminus
LMCA28.6%17.1%DeepSeek V3.1 Terminus
HellaSwagNot measured95.3%Only GPT-4 measured
WinoGrandeNot measured87.5%Only GPT-4 measured
LMArena Text14191206DeepSeek V3.1 Terminus
LMArena Text: instruction following14041201DeepSeek V3.1 Terminus
LMArena Text: creative writing14031191DeepSeek V3.1 Terminus
LMArena Text: English14291220DeepSeek V3.1 Terminus
LMArena Text: non-English14081194DeepSeek V3.1 Terminus
LMArena Text: ChineseNot measured1184Only GPT-4 measured
LMArena Text: GermanNot measured1198Only GPT-4 measured
LMArena Text: JapaneseNot measured1137Only GPT-4 measured
LMArena Text: KoreanNot measured1087Only GPT-4 measured
LMArena Text: FrenchNot measured1220Only GPT-4 measured
LMArena Text: SpanishNot measured1197Only GPT-4 measured
LMArena Text: Russian14351195DeepSeek V3.1 Terminus
LMArena Text: Polish14271265DeepSeek V3.1 Terminus
LMArena Text: multi-turn14111206DeepSeek V3.1 Terminus
LMArena Text: longer queries14221190DeepSeek V3.1 Terminus
LMArena Text: writing and language14091219DeepSeek V3.1 Terminus
LMArena Text: entertainment and media13911184DeepSeek V3.1 Terminus
Neither model has a result on the other 23 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for DeepSeek V3.1 Terminus and GPT-4
BenchmarkDeepSeek V3.1 TerminusGPT-4Leader
Neither model has a result on 5 boards here

What each workload costs

Three fixed token mixes at each model's listed price, so per-token rates become a bill.

Chat turn

1K input + 500 output tokens

DeepSeek V3.1 Terminus$0.0008Fits in one request
GPT-4$0.060Fits in one request

DeepSeek V3.1 Terminus costs less

Repository review

50K input + 3K output tokens

DeepSeek V3.1 Terminus$0.018Fits in one request
GPT-4$1.68Exceeds the context window

DeepSeek V3.1 Terminus costs less

Agent loop

200K input + 10K output tokens

DeepSeek V3.1 Terminus$0.071Exceeds the context window
GPT-4$6.60Exceeds the context window

DeepSeek V3.1 Terminus costs less

Listed OpenRouter rates, without cache or batch discounts. A workload that does not fit a model's context window would need splitting.

Specification differences

The leading value in a row is marked; a row with a missing value names no leader. Our two rows are marked in red.

Specifications of DeepSeek V3.1 Terminus and GPT-4
SpecificationDeepSeek V3.1 TerminusOpenAI GPT-4
MakerDeepSeekOpenAI
WeightsOpen weightsNo public weights listed
ReleasedSeptember 22, 2025May 28, 2023
Context window164K8K
Input $/Mtok$0.3063$30.00
Output $/Mtok$1.0142$60.00
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it00
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is DeepSeek V3.1 Terminus better than GPT-4 for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. DeepSeek V3.1 Terminus has the edge where output tokens cost 59.2x less, and where the context window is larger, 164K against 8K tokens.

Which is cheaper, DeepSeek V3.1 Terminus or GPT-4?

DeepSeek V3.1 Terminus costs $0.3063 per million input tokens and $1.0142 per million output tokens; GPT-4 costs $30.00 and $60.00. Cost per solved task is known for neither of them.

Which has the bigger context window, DeepSeek V3.1 Terminus or GPT-4?

DeepSeek V3.1 Terminus, with 164K tokens against 8K.

How many benchmarks is the DeepSeek V3.1 Terminus vs GPT-4 comparison based on?

27 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3.1 Terminus's coding index rests on 0 independent benchmarks and GPT-4's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next