Enter opens the highlighted result. Esc closes.

gpt-oss-120b vs gpt-oss-20b

gpt-oss-120b leads gpt-oss-20b on the Artificial Analysis Intelligence Index, 11.6 to 10.0, and gpt-oss-20b costs 3.3x less per output token ($0.15 vs $0.50 per million). Updated October 22, 2026; 62 benchmarks have a result for both.

Replace gpt-oss-120b Replace gpt-oss-20b
gpt-oss-120bOpenAI
Coding
No data
Overall
50.0
Output $/M
$0.50
61
gpt-oss-20bOpenAI
Coding
No data
Overall
38.5
Output $/M
$0.15
1
vs boards won of 62 shared

gpt-oss-120b or gpt-oss-20b: which to pick

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

Pick gpt-oss-120b for work measured by SciCode, where it leads.

Model Agpt-oss-120bOpenAINot benchmarked1 independent benchmark$0.51 per solved taskModel Bgpt-oss-20bOpenAINot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
62
gpt-oss-120b only
19
gpt-oss-20b only
0
Like-for-like categories
8 of 9

Only benchmarks that tested both models count toward a pick; a result only one model has is shown but not compared.

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
gpt-oss-120b not benchmarked, gpt-oss-20b not benchmarked.
Widest category
General, 12.3 points apart on average over 20 shared benchmarks.
Widest single gap
OTIS Mock AIME: gpt-oss-120b 88.9% against gpt-oss-20b 65.3%, 23.6 points apart on a 0 to 100 scale.
Output price
gpt-oss-120b $0.50 against gpt-oss-20b $0.15 per million output tokens, 3.3x apart.
Context window
Both take 131K tokens.

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
Categorygpt-oss-120bBarsgpt-oss-20bTies
Coding81
Intelligence20
Agentic work20
Reasoning80
Math80
Knowledge110
Web design20
General200
All shared boards611
CodingLike-for-likeSix of these boards feed the Vibe Coding Index81
Coding benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
Aider polyglot41.8%Not measuredOnly gpt-oss-120b measured
SWE-bench Verified26.0%Not measuredOnly gpt-oss-120b measured
AA Coding Index30.4Not measuredOnly gpt-oss-120b measured
LiveCodeBench (Vals AI)83.2%80.4%gpt-oss-120b
SWE-bench Verified (Vals AI)33.6%Not measuredOnly gpt-oss-120b measured
LiveCodeBench (Artificial Analysis run)87.8%77.7%gpt-oss-120b
SciCode34.0%38.9%gpt-oss-20b
Terminal-Bench Hard (Artificial Analysis run)23.5%10.6%gpt-oss-120b
Terminal-Bench18.7%3.4%gpt-oss-120b
WeirdML48.2%40.9%gpt-oss-120b
ALE-Bench575.6566.0gpt-oss-120b
LMArena Text: coding13791307gpt-oss-120b
LMArena Text: software and IT13871311gpt-oss-120b
SciCode7.7%Not measuredOnly gpt-oss-120b measured
Neither model has a result on the other 25 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index20
Intelligence benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
AA Intelligence Index11.610.0gpt-oss-120b
Epoch Capabilities Index139.8137.7gpt-oss-120b
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
Agentic work benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
AA Agentic Index3.7Not measuredOnly gpt-oss-120b measured
τ²-Bench Telecom65.8%60.2%gpt-oss-120b
APEX-Agents4.4%Not measuredOnly gpt-oss-120b measured
METR time horizons56.6%Not measuredOnly gpt-oss-120b measured
Vending-Bench 2-$22Not measuredOnly gpt-oss-120b measured
BrowseComp-Plus67.0%50.6%gpt-oss-120b
Neither model has a result on the other 35 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index80
Reasoning benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
GPQA Diamond78.5%68.9%gpt-oss-120b
GPQA Diamond (Artificial Analysis run)78.2%68.8%gpt-oss-120b
Humanity's Last Exam (Artificial Analysis run)19.6%11.0%gpt-oss-120b
GPQA Diamond (Epoch AI run)75.8%60.8%gpt-oss-120b
SimpleBench22.1%Not measuredOnly gpt-oss-120b measured
Mystery game puzzles2.0%Not measuredOnly gpt-oss-120b measured
LMArena Text: hard prompts13641274gpt-oss-120b
LMArena Text: expert13561258gpt-oss-120b
LMArena Text: hard prompts in English13681292gpt-oss-120b
Kagi LLM Benchmark58.6%53.2%gpt-oss-120b
Neither model has a result on the other 12 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index80
Math benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
AIME92.6%86.0%gpt-oss-120b
MATH 500 (Vals AI)94.8%94.2%gpt-oss-120b
MGSM (Vals AI)92.0%89.0%gpt-oss-120b
AA Math Index93.489.3gpt-oss-120b
AIME 2025 (Artificial Analysis run)93.4%89.3%gpt-oss-120b
OTIS Mock AIME88.9%65.3%gpt-oss-120b
LMArena Text: math13891317gpt-oss-120b
LMArena Text: mathematical work13861305gpt-oss-120b
Neither model has a result on the other 5 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index110
Knowledge benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
MMLU Pro79.2%71.6%gpt-oss-120b
LegalBench (Vals AI)75.9%70.8%gpt-oss-120b
CaseLaw (Vals AI)48.8%43.8%gpt-oss-120b
CorpFin (Vals AI)58.2%53.1%gpt-oss-120b
TaxEval (Vals AI)71.6%63.7%gpt-oss-120b
MedQA (Vals AI)91.4%82.9%gpt-oss-120b
MMLU Pro (Artificial Analysis run)80.8%74.8%gpt-oss-120b
LMArena Text: science13821296gpt-oss-120b
LMArena Text: medicine and healthcare13881310gpt-oss-120b
LMArena Text: legal and government13551280gpt-oss-120b
LMArena Text: business and finance13591284gpt-oss-120b
Neither model has a result on the other 7 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index20
Web design benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
Design Arena: websites967852gpt-oss-120b
Design Arena: UI components924Not measuredOnly gpt-oss-120b measured
Design Arena: game dev992Not measuredOnly gpt-oss-120b measured
Design Arena: data viz987933gpt-oss-120b
Design Arena: 3D901Not measuredOnly gpt-oss-120b measured
Design Arena: all code categories962Not measuredOnly gpt-oss-120b measured
Neither model has a result on the other 3 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index200
General benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
IFBench69.0%65.1%gpt-oss-120b
Long context reasoning (AA-LCR)52.0%34.7%gpt-oss-120b
DTBench76.3%68.0%gpt-oss-120b
LMCA22.1%14.5%gpt-oss-120b
Fiction.LiveBench44.4%Not measuredOnly gpt-oss-120b measured
Creative writing (Lech Mazur)77.1%Not measuredOnly gpt-oss-120b measured
Surface Evolver Bench25.0%Not measuredOnly gpt-oss-120b measured
LMArena Text13661288gpt-oss-120b
LMArena Text: instruction following13181236gpt-oss-120b
LMArena Text: creative writing12741201gpt-oss-120b
LMArena Text: English13751303gpt-oss-120b
LMArena Text: non-English13511269gpt-oss-120b
LMArena Text: Chinese13851315gpt-oss-120b
LMArena Text: German13531255gpt-oss-120b
LMArena Text: Japanese13311244gpt-oss-120b
LMArena Text: Korean12821236gpt-oss-120b
LMArena Text: French1370Not measuredOnly gpt-oss-120b measured
LMArena Text: Spanish13891267gpt-oss-120b
LMArena Text: Russian13431278gpt-oss-120b
LMArena Text: Polish13531252gpt-oss-120b
LMArena Text: multi-turn13401268gpt-oss-120b
LMArena Text: longer queries13191250gpt-oss-120b
LMArena Text: writing and language13061238gpt-oss-120b
LMArena Text: entertainment and media12921216gpt-oss-120b
Neither model has a result on the other 22 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for gpt-oss-120b and gpt-oss-20b
Benchmarkgpt-oss-120bgpt-oss-20bLeader
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

gpt-oss-120b$0.0003Fits in one request
gpt-oss-20b$0.0001Fits in one request

gpt-oss-20b costs less

Repository review

50K input + 3K output tokens

gpt-oss-120b$0.0065Fits in one request
gpt-oss-20b$0.0019Fits in one request

gpt-oss-20b costs less

Agent loop

200K input + 10K output tokens

gpt-oss-120b$0.025Exceeds the context window
gpt-oss-20b$0.0075Exceeds the context window

gpt-oss-20b 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 gpt-oss-120b and gpt-oss-20b
SpecificationOpenAI gpt-oss-120bOpenAI gpt-oss-20b
MakerOpenAIOpenAI
WeightsOpen weightsOpen weights
ReleasedAugust 5, 2025August 5, 2025
Context window131K131K
Input $/Mtok$0.10$0.03
Output $/Mtok$0.50$0.15
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it10
Cost per solved task$0.51Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is gpt-oss-120b better than gpt-oss-20b for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick gpt-oss-120b for work measured by SciCode, where it leads.

Which is cheaper, gpt-oss-120b or gpt-oss-20b?

gpt-oss-120b costs $0.10 per million input tokens and $0.50 per million output tokens; gpt-oss-20b costs $0.03 and $0.15. Cost per solved task is known for gpt-oss-120b only.

Which has the bigger context window, gpt-oss-120b or gpt-oss-20b?

Neither: both take 131K tokens.

How many benchmarks is the gpt-oss-120b vs gpt-oss-20b comparison based on?

62 benchmarks have a result for both models as of October 22, 2026. gpt-oss-120b's coding index rests on 1 independent benchmark and gpt-oss-20b's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next