Enter opens the highlighted result. Esc closes.

Llama 3.3 70B Instruct vs Llama 4 Scout

Llama 3.3 70B Instruct leads Llama 4 Scout on the Artificial Analysis Intelligence Index, 7.7 to 6.5, and Llama 4 Scout costs 1.4x less per output token ($0.50 vs $0.72 per million). Updated October 21, 2026; 48 benchmarks have a result for both.

Replace Llama 3.3 70B Instruct Replace Llama 4 Scout
Llama 3.3 70B InstructMeta Llama
Coding
No data
Overall
26.0
Output $/M
$0.72
11
Llama 4 ScoutMeta Llama
Coding
No data
Overall
24.9
Output $/M
$0.50
37
vs boards won of 48 shared

Llama 3.3 70B Instruct or Llama 4 Scout: which to pick

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

Pick Llama 3.3 70B Instruct for work measured by LiveCodeBench (Artificial Analysis run) or LMArena Text: coding, where it leads.

Model ALlama 3.3 70B InstructMeta LlamaNot benchmarkedNo coding benchmarkNo cost per solved taskModel BLlama 4 ScoutMeta LlamaNot benchmarkedNo coding benchmark$1.276 per solved task
Shared results
48
Llama 3.3 70B Instruct only
4
Llama 4 Scout only
34
Like-for-like categories
7 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
Llama 3.3 70B Instruct not benchmarked, Llama 4 Scout not benchmarked.
Widest category
Agentic work, 7.4 points apart on average over 2 shared benchmarks.
Widest single gap
MATH level 5: Llama 3.3 70B Instruct 41.6% against Llama 4 Scout 62.3%, 20.7 points apart on a 0 to 100 scale.
Output price
Llama 3.3 70B Instruct $0.72 against Llama 4 Scout $0.50 per million output tokens, 1.4x apart.
Context window
Llama 3.3 70B Instruct takes 131K tokens, Llama 4 Scout 1.31M.

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
CategoryLlama 3.3 70B InstructBarsLlama 4 ScoutTies
Coding23
Intelligence11
Agentic work20
Reasoning06
Math06
Knowledge23
General418
All shared boards1137
CodingLike-for-likeSix of these boards feed the Vibe Coding Index23
Coding benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
SWE-bench VerifiedNot measured9.1%Only Llama 4 Scout measured
AA Coding Index11.98.2Llama 3.3 70B Instruct
LiveCodeBench (Vals AI)Not measured38.5%Only Llama 4 Scout measured
LiveCodeBench (Artificial Analysis run)28.8%29.9%Llama 4 Scout
SciCodeNot measured21.3%Only Llama 4 Scout measured
Terminal-Bench Hard (Artificial Analysis run)3.0%1.5%Llama 3.3 70B Instruct
WeirdML14.4%Not measuredOnly Llama 3.3 70B Instruct measured
LMArena Text: coding12681286Llama 4 Scout
LMArena Text: software and IT12731286Llama 4 Scout
Neither model has a result on the other 30 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index11
Intelligence benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
AA Intelligence Index7.76.5Llama 3.3 70B Instruct
Epoch Capabilities Index127.3129.6Llama 4 Scout
Agentic workLike-for-likePublic boards, shown as published, kept out of the index20
Agentic work benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
AA Agentic IndexNot measured0.5Only Llama 4 Scout measured
τ²-Bench Telecom26.6%15.5%Llama 3.3 70B Instruct
BALROG23.0%Not measuredOnly Llama 3.3 70B Instruct measured
Berkeley Function Calling Leaderboard (BFCL)31.9%28.1%Llama 3.3 70B Instruct
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index06
Reasoning benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
GPQA DiamondNot measured47.0%Only Llama 4 Scout measured
MMMU Pro (Vals AI)Not measured58.8%Only Llama 4 Scout measured
GPQA Diamond (Artificial Analysis run)49.8%58.7%Llama 4 Scout
Humanity's Last Exam (Artificial Analysis run)3.6%3.8%Llama 4 Scout
GPQA Diamond (Epoch AI run)47.4%51.8%Llama 4 Scout
ARC-AGINot measured0.5%Only Llama 4 Scout measured
ARC-AGI-2Not measured0.0%Only Llama 4 Scout measured
SimpleBench19.9%Not measuredOnly Llama 3.3 70B Instruct measured
LMArena Text: hard prompts12571266Llama 4 Scout
LMArena Text: expert12251235Llama 4 Scout
LMArena Text: hard prompts in English12771293Llama 4 Scout
LMArena Vision: diagramsNot measured1125Only Llama 4 Scout measured
LMArena Vision: homeworkNot measured1148Only Llama 4 Scout measured
Kagi LLM BenchmarkNot measured36.9%Only Llama 4 Scout measured
Neither model has a result on the other 8 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index06
Math benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
AIMENot measured19.0%Only Llama 4 Scout measured
MATH 500 (Vals AI)Not measured79.2%Only Llama 4 Scout measured
MGSM (Vals AI)Not measured88.0%Only Llama 4 Scout measured
AA Math Index7.714.0Llama 4 Scout
AIME 2025 (Artificial Analysis run)7.7%14.0%Llama 4 Scout
OTIS Mock AIME5.1%7.8%Llama 4 Scout
MATH level 541.6%62.3%Llama 4 Scout
LMArena Text: math12671287Llama 4 Scout
LMArena Text: mathematical work12671286Llama 4 Scout
Neither model has a result on the other 4 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index23
Knowledge benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
MMLU ProNot measured69.6%Only Llama 4 Scout measured
LegalBench (Vals AI)Not measured72.0%Only Llama 4 Scout measured
CorpFin (Vals AI)Not measured46.8%Only Llama 4 Scout measured
TaxEval (Vals AI)Not measured55.2%Only Llama 4 Scout measured
MortgageTax (Vals AI)Not measured57.8%Only Llama 4 Scout measured
MedQA (Vals AI)Not measured50.9%Only Llama 4 Scout measured
MedCode (Vals AI)Not measured23.3%Only Llama 4 Scout measured
MMLU Pro (Artificial Analysis run)71.3%75.2%Llama 4 Scout
MMLU86.3%Not measuredOnly Llama 3.3 70B Instruct measured
LMArena Text: science12841278Llama 3.3 70B Instruct
LMArena Text: medicine and healthcare12751275Llama 3.3 70B Instruct
LMArena Text: legal and government12841285Llama 4 Scout
LMArena Text: business and finance12431257Llama 4 Scout
Neither model has a result on the other 5 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index5 boards, none shared
Web design benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
Design Arena: websitesNot measured750Only Llama 4 Scout measured
Design Arena: UI componentsNot measured768Only Llama 4 Scout measured
Design Arena: game devNot measured773Only Llama 4 Scout measured
Design Arena: data vizNot measured896Only Llama 4 Scout measured
Design Arena: all code categoriesNot measured790Only Llama 4 Scout measured
Neither model has a result on the other 4 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index418
General benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
SAGE (Vals AI)Not measured34.8%Only Llama 4 Scout measured
MedScribe (Vals AI)Not measured50.6%Only Llama 4 Scout measured
IFBench47.1%39.5%Llama 3.3 70B Instruct
Long context reasoning (AA-LCR)15.7%27.7%Llama 4 Scout
DTBench59.5%57.9%Llama 3.3 70B Instruct
LMCA17.5%12.0%Llama 3.3 70B Instruct
Fiction.LiveBench33.3%36.0%Llama 4 Scout
LMArena Text12741279Llama 4 Scout
LMArena Text: instruction following12421248Llama 4 Scout
LMArena Text: creative writing12501249Llama 3.3 70B Instruct
LMArena VisionNot measured1118Only Llama 4 Scout measured
LMArena Text: English13031304Llama 4 Scout
LMArena Text: non-English12361252Llama 4 Scout
LMArena Text: Chinese12171255Llama 4 Scout
LMArena Text: German12511273Llama 4 Scout
LMArena Text: Japanese11501206Llama 4 Scout
LMArena Text: Korean11421207Llama 4 Scout
LMArena Text: French12821283Llama 4 Scout
LMArena Text: Spanish12691278Llama 4 Scout
LMArena Text: Russian12521263Llama 4 Scout
LMArena Text: Polish12591265Llama 4 Scout
LMArena Text: multi-turn12801280Llama 4 Scout
LMArena Text: longer queries12551265Llama 4 Scout
LMArena Text: writing and language12551269Llama 4 Scout
LMArena Text: entertainment and media12411242Llama 4 Scout
LMArena Vision: EnglishNot measured1143Only Llama 4 Scout measured
LMArena Vision: ChineseNot measured1106Only Llama 4 Scout measured
LMArena Vision: reading textNot measured1117Only Llama 4 Scout measured
LMArena Vision: humorNot measured1102Only Llama 4 Scout measured
LMArena Vision: creative writingNot measured1126Only Llama 4 Scout measured
Neither model has a result on the other 16 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Llama 3.3 70B Instruct and Llama 4 Scout
BenchmarkLlama 3.3 70B InstructLlama 4 ScoutLeader
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

Llama 3.3 70B Instruct$0.0008Fits in one request
Llama 4 Scout$0.0004Fits in one request

Llama 4 Scout costs less

Repository review

50K input + 3K output tokens

Llama 3.3 70B Instruct$0.025Fits in one request
Llama 4 Scout$0.010Fits in one request

Llama 4 Scout costs less

Agent loop

200K input + 10K output tokens

Llama 3.3 70B Instruct$0.097Exceeds the context window
Llama 4 Scout$0.040Fits in one request

Llama 4 Scout 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 Llama 3.3 70B Instruct and Llama 4 Scout
SpecificationMeta Llama 3.3 70B InstructMeta Llama 4 Scout
MakerMeta LlamaMeta Llama
WeightsOpen weightsOpen weights
ReleasedDecember 6, 2024April 5, 2025
Context window131K1.31M
Input $/Mtok$0.45$0.175
Output $/Mtok$0.72$0.50
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it00
Cost per solved taskNot computable$1.276
SourceOpenRouter listingOpenRouter listing

Questions

Is Llama 3.3 70B Instruct better than Llama 4 Scout for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Llama 3.3 70B Instruct for work measured by LiveCodeBench (Artificial Analysis run) or LMArena Text: coding, where it leads.

Which is cheaper, Llama 3.3 70B Instruct or Llama 4 Scout?

Llama 3.3 70B Instruct costs $0.45 per million input tokens and $0.72 per million output tokens; Llama 4 Scout costs $0.175 and $0.50. Cost per solved task is known for Llama 4 Scout only.

Which has the bigger context window, Llama 3.3 70B Instruct or Llama 4 Scout?

Llama 4 Scout, with 1.31M tokens against 131K.

How many benchmarks is the Llama 3.3 70B Instruct vs Llama 4 Scout comparison based on?

48 benchmarks have a result for both models as of October 21, 2026. Llama 3.3 70B Instruct's coding index rests on 0 independent benchmarks and Llama 4 Scout's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next