Enter opens the highlighted result. Esc closes.

Phi 4 vs Qwen3 30B A3B Instruct 2507

Qwen3 30B A3B Instruct 2507 leads Phi 4 on the Artificial Analysis Intelligence Index, 7.5 to 5.9, and Phi 4 costs 2.1x less per output token ($0.14 vs $0.30 per million). Updated October 21, 2026; 40 benchmarks have a result for both.

Replace Phi 4 Replace Qwen3 30B A3B Instruct 2507
Phi 4Microsoft
Coding
No data
Overall
5.9
Output $/M
$0.14
1
Qwen3 30B A3B Instruct 2507Qwen
Coding
No data
Overall
7.5
Output $/M
$0.30
39
vs boards won of 40 shared

Phi 4 or Qwen3 30B A3B Instruct 2507: which to pick

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

Pick Phi 4 for output tokens 2.1x cheaper ($0.14 vs $0.30 per million) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.

Model APhi 4MicrosoftNot benchmarkedNo coding benchmarkNo cost per solved taskModel BQwen3 30B A3B Instruct 2507QwenNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
40
Phi 4 only
5
Qwen3 30B A3B Instruct 2507 only
4
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
Phi 4 not benchmarked, Qwen3 30B A3B Instruct 2507 not benchmarked.
Widest category
Math, 39.2 points apart on average over 5 shared benchmarks.
Widest single gap
Long context reasoning (AA-LCR): Phi 4 0.0% against Qwen3 30B A3B Instruct 2507 61.3%, 61.3 points apart on a 0 to 100 scale.
Output price
Phi 4 $0.14 against Qwen3 30B A3B Instruct 2507 $0.30 per million output tokens, 2.1x apart.
Context window
Phi 4 takes 16K tokens, Qwen3 30B A3B Instruct 2507 262K.

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
CategoryPhi 4BarsQwen3 30B A3B Instruct 2507Ties
Coding04
Intelligence01
Agentic work01
Reasoning15
Math05
Knowledge05
General018
All shared boards139
CodingLike-for-likeSix of these boards feed the Vibe Coding Index04
Coding benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
LiveCodeBench (Artificial Analysis run)23.1%70.7%Qwen3 30B A3B Instruct 2507
SciCodeNot measured33.0%Only Qwen3 30B A3B Instruct 2507 measured
Terminal-Bench Hard (Artificial Analysis run)3.8%6.1%Qwen3 30B A3B Instruct 2507
LMArena Text: coding12311416Qwen3 30B A3B Instruct 2507
LMArena Text: software and IT12281417Qwen3 30B A3B Instruct 2507
Neither model has a result on the other 33 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
Intelligence benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
AA Intelligence Index5.97.5Qwen3 30B A3B Instruct 2507
Epoch Capabilities Index130.5Not measuredOnly Phi 4 measured
Agentic workLike-for-likePublic boards, shown as published, kept out of the index01
Agentic work benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
τ²-Bench Telecom0.0%28.1%Qwen3 30B A3B Instruct 2507
BALROG11.6%Not measuredOnly Phi 4 measured
Neither model has a result on the other 37 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index15
Reasoning benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
GPQA Diamond (Artificial Analysis run)57.5%70.7%Qwen3 30B A3B Instruct 2507
Humanity's Last Exam (Artificial Analysis run)3.8%10.3%Qwen3 30B A3B Instruct 2507
GPQA Diamond (Epoch AI run)56.1%55.6%Phi 4
LMArena Text: hard prompts12201398Qwen3 30B A3B Instruct 2507
LMArena Text: expert12031397Qwen3 30B A3B Instruct 2507
LMArena Text: hard prompts in English12301406Qwen3 30B A3B Instruct 2507
Neither model has a result on the other 15 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index05
Math benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
AA Math Index18.066.3Qwen3 30B A3B Instruct 2507
AIME 2025 (Artificial Analysis run)18.0%66.3%Qwen3 30B A3B Instruct 2507
OTIS Mock AIME13.8%62.2%Qwen3 30B A3B Instruct 2507
MATH level 564.9%Not measuredOnly Phi 4 measured
LMArena Text: math12461394Qwen3 30B A3B Instruct 2507
LMArena Text: mathematical work12541411Qwen3 30B A3B Instruct 2507
Neither model has a result on the other 7 boards here
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index05
Knowledge benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
MMLU Pro (Artificial Analysis run)71.4%80.5%Qwen3 30B A3B Instruct 2507
MMLU84.8%Not measuredOnly Phi 4 measured
LMArena Text: science12061393Qwen3 30B A3B Instruct 2507
LMArena Text: medicine and healthcare12021392Qwen3 30B A3B Instruct 2507
LMArena Text: legal and government12411387Qwen3 30B A3B Instruct 2507
LMArena Text: business and finance11991389Qwen3 30B A3B Instruct 2507
Neither model has a result on the other 11 boards here
Web designNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Web design benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
Neither model has a result on 9 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index018
General benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
IFBench23.5%50.7%Qwen3 30B A3B Instruct 2507
Long context reasoning (AA-LCR)0.0%61.3%Qwen3 30B A3B Instruct 2507
DTBenchNot measured67.2%Only Qwen3 30B A3B Instruct 2507 measured
LMCANot measured22.4%Only Qwen3 30B A3B Instruct 2507 measured
Creative writing (Lech Mazur)62.6%Not measuredOnly Phi 4 measured
LMArena Text12171384Qwen3 30B A3B Instruct 2507
LMArena Text: instruction following12011363Qwen3 30B A3B Instruct 2507
LMArena Text: creative writing11811317Qwen3 30B A3B Instruct 2507
LMArena Text: English12311391Qwen3 30B A3B Instruct 2507
LMArena Text: non-English11971372Qwen3 30B A3B Instruct 2507
LMArena Text: Chinese12121432Qwen3 30B A3B Instruct 2507
LMArena Text: German12221379Qwen3 30B A3B Instruct 2507
LMArena Text: Japanese11581337Qwen3 30B A3B Instruct 2507
LMArena Text: Korean11511330Qwen3 30B A3B Instruct 2507
LMArena Text: French12241419Qwen3 30B A3B Instruct 2507
LMArena Text: Spanish12341404Qwen3 30B A3B Instruct 2507
LMArena Text: Russian12091369Qwen3 30B A3B Instruct 2507
LMArena Text: PolishNot measured1353Only Qwen3 30B A3B Instruct 2507 measured
LMArena Text: multi-turn12061378Qwen3 30B A3B Instruct 2507
LMArena Text: longer queries12171378Qwen3 30B A3B Instruct 2507
LMArena Text: writing and language11981336Qwen3 30B A3B Instruct 2507
LMArena Text: entertainment and media11621334Qwen3 30B A3B Instruct 2507
Neither model has a result on the other 24 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for Phi 4 and Qwen3 30B A3B Instruct 2507
BenchmarkPhi 4Qwen3 30B A3B Instruct 2507Leader
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

Phi 4$0.0001Fits in one request
Qwen3 30B A3B Instruct 2507$0.0002Fits in one request

Phi 4 costs less

Repository review

50K input + 3K output tokens

Phi 4$0.0039Exceeds the context window
Qwen3 30B A3B Instruct 2507$0.0054Fits in one request

Phi 4 costs less

Agent loop

200K input + 10K output tokens

Phi 4$0.015Exceeds the context window
Qwen3 30B A3B Instruct 2507$0.021Fits in one request

Phi 4 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 Phi 4 and Qwen3 30B A3B Instruct 2507
SpecificationMicrosoft Phi 4Qwen3 30B A3B Instruct 2507
MakerMicrosoftQwen
WeightsOpen weightsOpen weights
ReleasedJanuary 10, 2025July 29, 2025
Context window16K262K
Input $/Mtok$0.07$0.09
Output $/Mtok$0.14$0.30
Coding indexNot benchmarkedNot benchmarked
Independent measurements behind it00
Cost per solved taskNot computableNot computable
SourceOpenRouter listingOpenRouter listing

Questions

Is Phi 4 better than Qwen3 30B A3B Instruct 2507 for coding?

Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Phi 4 for output tokens 2.1x cheaper ($0.14 vs $0.30 per million) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.

Which is cheaper, Phi 4 or Qwen3 30B A3B Instruct 2507?

Phi 4 costs $0.07 per million input tokens and $0.14 per million output tokens; Qwen3 30B A3B Instruct 2507 costs $0.09 and $0.30. Cost per solved task is known for neither of them.

Which has the bigger context window, Phi 4 or Qwen3 30B A3B Instruct 2507?

Qwen3 30B A3B Instruct 2507, with 262K tokens against 16K.

How many benchmarks is the Phi 4 vs Qwen3 30B A3B Instruct 2507 comparison based on?

40 benchmarks have a result for both models as of October 21, 2026. Phi 4's coding index rests on 0 independent benchmarks and Qwen3 30B A3B Instruct 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next