Enter opens the highlighted result. Esc closes.

gpt-oss-20b vs Grok 4.3

Grok 4.3 leads gpt-oss-20b on the Artificial Analysis Intelligence Index, 24.9 to 10.0, and gpt-oss-20b costs 16.7x less per output token ($0.15 vs $2.50 per million). Updated October 21, 2026; 50 benchmarks have a result for both.

Replace gpt-oss-20b Replace Grok 4.3
gpt-oss-20bOpenAI
Coding
No data
Overall
38.5
Output $/M
$0.15
0
Grok 4.3xAI
Coding
37.4
Overall
53.1
Output $/M
$2.50
50
vs boards won of 50 shared

gpt-oss-20b or Grok 4.3: which to pick

Only Grok 4.3 has a coding score, 37.4 on the Vibe Coding Index; gpt-oss-20b has none, so the two cannot be ranked on coding.

Pick gpt-oss-20b for output tokens 16.7x cheaper ($0.15 vs $2.50 per million).

Model Agpt-oss-20bOpenAINot benchmarkedNo coding benchmarkNo cost per solved taskModel BGrok 4.3xAI37.4/1003 independent benchmarks$0.31 per solved task
Shared results
50
gpt-oss-20b only
12
Grok 4.3 only
54
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-20b not benchmarked, Grok 4.3 37.4 from 3 benchmarks.
Widest category
Agentic work, 37.4 points apart on average over 1 shared benchmark.
Widest single gap
Design Arena: websites: gpt-oss-20b 852 against Grok 4.3 1196, 43.0 points apart on a 0 to 100 scale.
Output price
gpt-oss-20b $0.15 against Grok 4.3 $2.50 per million output tokens, 16.7x apart.
Context window
gpt-oss-20b takes 131K tokens, Grok 4.3 1M.

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-20bBarsGrok 4.3Ties
Coding07
Intelligence01
Agentic work01
Reasoning07
Math03
Knowledge09
Web design02
General020
All shared boards050
CodingLike-for-likeSix of these boards feed the Vibe Coding Index07
Coding benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
LMArena WebDevNot measured1358Only Grok 4.3 measured
AA Coding IndexNot measured42.2Only Grok 4.3 measured
LiveCodeBench (Vals AI)80.4%84.5%Grok 4.3
SWE-bench Verified (Vals AI)Not measured71.4%Only Grok 4.3 measured
Vibe Code Bench (Vals AI)Not measured19.4%Only Grok 4.3 measured
Code migration (Vals AI)Not measured6.8%Only Grok 4.3 measured
ProgramBench (Vals AI)Not measured0.0%Only Grok 4.3 measured
Terminal-Bench 2.1 (Vals AI)Not measured42.0%Only Grok 4.3 measured
SkillsBench (Vals AI)Not measured40.6%Only Grok 4.3 measured
LiveCodeBench (Artificial Analysis run)77.7%Not measuredOnly gpt-oss-20b measured
SciCode38.9%48.3%Grok 4.3
Terminal-Bench Hard (Artificial Analysis run)10.6%37.9%Grok 4.3
Terminal-Bench3.4%Not measuredOnly gpt-oss-20b measured
WeirdML40.9%49.9%Grok 4.3
ALE-Bench566.0944.2Grok 4.3
LMArena Text: coding13071415Grok 4.3
LMArena Text: software and IT13111416Grok 4.3
LMArena WebDev: ReactNot measured1353Only Grok 4.3 measured
LMArena WebDev: HTMLNot measured1367Only Grok 4.3 measured
LMArena WebDev: from an imageNot measured1375Only Grok 4.3 measured
Neither model has a result on the other 19 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
Intelligence benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
AA Intelligence Index10.024.9Grok 4.3
Epoch Capabilities Index137.7Not measuredOnly gpt-oss-20b measured
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
AA Agentic IndexNot measured15.5Only Grok 4.3 measured
Design Arena: agents building web appsNot measured1126Only Grok 4.3 measured
Design Arena: agents building full-stack appsNot measured1006Only Grok 4.3 measured
Design Arena: agents building mobile appsNot measured1059Only Grok 4.3 measured
Design Arena: agents building Android appsNot measured961Only Grok 4.3 measured
Design Arena: agents building Godot gamesNot measured914Only Grok 4.3 measured
Design Arena: agents building gamesNot measured1007Only Grok 4.3 measured
Design Arena: HTML slidesNot measured1008Only Grok 4.3 measured
Design Arena: PowerPoint slidesNot measured1013Only Grok 4.3 measured
Finance Agent (Vals AI)Not measured37.7%Only Grok 4.3 measured
Tax Agent Bench (Vals AI)Not measured6.2%Only Grok 4.3 measured
Legal Research Bench (Vals AI)Not measured15.4%Only Grok 4.3 measured
Excel Modeling Benchmark (Vals AI)Not measured18.4%Only Grok 4.3 measured
τ²-Bench Telecom60.2%97.7%Grok 4.3
Vending-Bench 2Not measured$35Only Grok 4.3 measured
LMArena SearchNot measured1166Only Grok 4.3 measured
BrowseComp-Plus50.6%Not measuredOnly gpt-oss-20b measured
Neither model has a result on the other 24 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index07
Reasoning benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
GPQA Diamond68.9%91.4%Grok 4.3
MMMU Pro (Vals AI)Not measured83.1%Only Grok 4.3 measured
GPQA Diamond (Artificial Analysis run)68.8%90.1%Grok 4.3
Humanity's Last Exam (Artificial Analysis run)11.0%37.2%Grok 4.3
GPQA Diamond (Epoch AI run)60.8%88.8%Grok 4.3
Chess puzzlesNot measured18.0%Only Grok 4.3 measured
LMArena Text: hard prompts12741396Grok 4.3
LMArena Text: expert12581384Grok 4.3
LMArena Text: hard prompts in English12921402Grok 4.3
LMArena Vision: diagramsNot measured1241Only Grok 4.3 measured
LMArena Vision: homeworkNot measured1241Only Grok 4.3 measured
Kagi LLM Benchmark53.2%Not measuredOnly gpt-oss-20b measured
Neither model has a result on the other 10 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index03
Math benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
AIME86.0%Not measuredOnly gpt-oss-20b measured
MATH 500 (Vals AI)94.2%Not measuredOnly gpt-oss-20b measured
MGSM (Vals AI)89.0%Not measuredOnly gpt-oss-20b measured
AA Math Index89.3Not measuredOnly gpt-oss-20b measured
AIME 2025 (Artificial Analysis run)89.3%Not measuredOnly gpt-oss-20b measured
OTIS Mock AIME65.3%93.3%Grok 4.3
FrontierMath, tiers 1 to 3Not measured42.8%Only Grok 4.3 measured
FrontierMath, tier 4Not measured14.6%Only Grok 4.3 measured
ProofBenchNot measured11.0%Only Grok 4.3 measured
LMArena Text: math13171387Grok 4.3
LMArena Text: mathematical work13051396Grok 4.3
Neither model has a result on the other 2 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index09
Knowledge benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
MMLU Pro71.6%85.8%Grok 4.3
LegalBench (Vals AI)70.8%84.5%Grok 4.3
CaseLaw (Vals AI)43.8%79.3%Grok 4.3
CorpFin (Vals AI)53.1%68.5%Grok 4.3
TaxEval (Vals AI)63.7%70.8%Grok 4.3
MortgageTax (Vals AI)Not measured48.3%Only Grok 4.3 measured
MedQA (Vals AI)82.9%Not measuredOnly gpt-oss-20b measured
MedCode (Vals AI)Not measured38.1%Only Grok 4.3 measured
MMLU Pro (Artificial Analysis run)74.8%Not measuredOnly gpt-oss-20b measured
SimpleQA VerifiedNot measured33.2%Only Grok 4.3 measured
LMArena Text: science12961399Grok 4.3
LMArena Text: medicine and healthcare13101384Grok 4.3
LMArena Text: legal and government12801386Grok 4.3
LMArena Text: business and finance12841381Grok 4.3
Neither model has a result on the other 4 boards here
Web designLike-for-likePublic boards, shown as published, kept out of the index02
Web design benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
Design Arena: websites8521196Grok 4.3
Design Arena: UI componentsNot measured1192Only Grok 4.3 measured
Design Arena: game devNot measured1174Only Grok 4.3 measured
Design Arena: data viz9331179Grok 4.3
Design Arena: 3DNot measured1125Only Grok 4.3 measured
Design Arena: all code categoriesNot measured1187Only Grok 4.3 measured
Design Arena: SVGNot measured1081Only Grok 4.3 measured
Design Arena: ASCII artNot measured1125Only Grok 4.3 measured
Neither model has a result on the other 1 board here
GeneralLike-for-likePublic boards, shown as published, kept out of the index020
General benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
Public Benefits Bench (Vals AI)Not measured51.7%Only Grok 4.3 measured
SAGE (Vals AI)Not measured19.7%Only Grok 4.3 measured
MedScribe (Vals AI)Not measured74.4%Only Grok 4.3 measured
Vals Multimodal IndexNot measured43.3%Only Grok 4.3 measured
IFBench65.1%83.3%Grok 4.3
Long context reasoning (AA-LCR)34.7%75.0%Grok 4.3
DTBench68.0%90.7%Grok 4.3
LMCA14.5%38.3%Grok 4.3
LMArena Text12881397Grok 4.3
LMArena Text: instruction following12361366Grok 4.3
LMArena Text: creative writing12011380Grok 4.3
LMArena VisionNot measured1229Only Grok 4.3 measured
LMArena Text: English13031404Grok 4.3
LMArena Text: non-English12691385Grok 4.3
LMArena Text: Chinese13151422Grok 4.3
LMArena Text: German12551395Grok 4.3
LMArena Text: Japanese12441379Grok 4.3
LMArena Text: Korean12361354Grok 4.3
LMArena Text: FrenchNot measured1411Only Grok 4.3 measured
LMArena Text: Spanish12671399Grok 4.3
LMArena Text: Russian12781399Grok 4.3
LMArena Text: Polish12521398Grok 4.3
LMArena Text: multi-turn12681406Grok 4.3
LMArena Text: longer queries12501393Grok 4.3
LMArena Text: writing and language12381385Grok 4.3
LMArena Text: entertainment and media12161370Grok 4.3
LMArena Vision: EnglishNot measured1235Only Grok 4.3 measured
LMArena Vision: ChineseNot measured1257Only Grok 4.3 measured
LMArena Vision: reading textNot measured1235Only Grok 4.3 measured
LMArena Vision: humorNot measured1222Only Grok 4.3 measured
LMArena Vision: recognizing thingsNot measured1217Only Grok 4.3 measured
LMArena Vision: creative writingNot measured1237Only Grok 4.3 measured
Neither model has a result on the other 14 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for gpt-oss-20b and Grok 4.3
Benchmarkgpt-oss-20bGrok 4.3Leader
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-20b$0.0001Fits in one request
Grok 4.3$0.0025Fits in one request

gpt-oss-20b costs less

Repository review

50K input + 3K output tokens

gpt-oss-20b$0.0019Fits in one request
Grok 4.3$0.070Fits in one request

gpt-oss-20b costs less

Agent loop

200K input + 10K output tokens

gpt-oss-20b$0.0075Exceeds the context window
Grok 4.3$0.275Fits in one request

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-20b and Grok 4.3
SpecificationOpenAI gpt-oss-20bSpaceXAI Grok 4.3
MakerOpenAIxAI
WeightsOpen weightsNo public weights listed
ReleasedAugust 5, 2025April 30, 2026
Context window131K1M
Input $/Mtok$0.03$1.25
Output $/Mtok$0.15$2.50
Coding indexNot benchmarked37.4
Independent measurements behind it03
Cost per solved taskNot computable$0.31
SourceOpenRouter listingOpenRouter listing

Questions

Is gpt-oss-20b better than Grok 4.3 for coding?

Only Grok 4.3 has a coding score, 37.4 on the Vibe Coding Index; gpt-oss-20b has none, so the two cannot be ranked on coding. Pick gpt-oss-20b for output tokens 16.7x cheaper ($0.15 vs $2.50 per million).

Which is cheaper, gpt-oss-20b or Grok 4.3?

gpt-oss-20b costs $0.03 per million input tokens and $0.15 per million output tokens; Grok 4.3 costs $1.25 and $2.50. Cost per solved task is known for Grok 4.3 only.

Which has the bigger context window, gpt-oss-20b or Grok 4.3?

Grok 4.3, with 1M tokens against 131K.

How many benchmarks is the gpt-oss-20b vs Grok 4.3 comparison based on?

50 benchmarks have a result for both models as of October 21, 2026. gpt-oss-20b's coding index rests on 0 independent benchmarks and Grok 4.3's on 3. Where only one model has a result, the page says so instead of naming a winner.

Compare next