Enter opens the highlighted result. Esc closes.

GPT-5.4 Nano vs Qwen3 235B A22B Thinking 2507

GPT-5.4 Nano leads Qwen3 235B A22B Thinking 2507 on the Artificial Analysis Intelligence Index, 20.7 to 12.7, and GPT-5.4 Nano costs 2.8x less per output token ($1.25 vs $3.50 per million). Updated October 22, 2026; 37 benchmarks have a result for both.

Replace GPT-5.4 Nano Replace Qwen3 235B A22B Thinking 2507
GPT-5.4 NanoOpenAI
Coding
44.8
Overall
45.7
Output $/M
$1.25
6
Qwen3 235B A22B Thinking 2507Qwen
Coding
No data
Overall
52.9
Output $/M
$3.50
29
vs boards won of 37 shared, 2 tied

GPT-5.4 Nano or Qwen3 235B A22B Thinking 2507: which to pick

Only GPT-5.4 Nano has a coding score, 44.8 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding.

Pick Qwen3 235B A22B Thinking 2507 for work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Model AGPT-5.4 NanoOpenAI44.8/1002 independent benchmarks$0.14 per solved taskModel BQwen3 235B A22B Thinking 2507QwenNot benchmarkedNo coding benchmarkNo cost per solved task
Shared results
37
GPT-5.4 Nano only
48
Qwen3 235B A22B Thinking 2507 only
9
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
GPT-5.4 Nano 44.8 from 2 benchmarks, Qwen3 235B A22B Thinking 2507 not benchmarked.
Widest category
Agentic work, 14.7 points apart on average over 1 shared benchmark.
Widest single gap
AA Coding Index: GPT-5.4 Nano 56.1 against Qwen3 235B A22B Thinking 2507 22.1, 34.0 points apart on a 0 to 100 scale.
Output price
GPT-5.4 Nano $1.25 against Qwen3 235B A22B Thinking 2507 $3.50 per million output tokens, 2.8x apart.
Context window
GPT-5.4 Nano takes 400K tokens, Qwen3 235B A22B Thinking 2507 128K.

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-5.4 NanoBarsQwen3 235B A22B Thinking 2507Ties
Coding22
Intelligence10
Agentic work10
Reasoning041 tie
Math12
Knowledge05
General1161 tie
All shared boards6292 ties
CodingLike-for-likeSix of these boards feed the Vibe Coding Index22
Coding benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
AA Coding Index56.122.1GPT-5.4 Nano
LiveCodeBench (Vals AI)84.0%Not measuredOnly GPT-5.4 Nano measured
SWE-bench Verified (Vals AI)69.8%Not measuredOnly GPT-5.4 Nano measured
Vibe Code Bench (Vals AI)26.1%Not measuredOnly GPT-5.4 Nano measured
Code migration (Vals AI)14.5%Not measuredOnly GPT-5.4 Nano measured
Terminal-Bench 2.1 (Vals AI)41.6%Not measuredOnly GPT-5.4 Nano measured
SciCode47.2%Not measuredOnly GPT-5.4 Nano measured
Terminal-Bench Hard (Artificial Analysis run)42.4%Not measuredOnly GPT-5.4 Nano measured
WeirdML49.2%41.0%GPT-5.4 Nano
ALE-Bench1004.5Not measuredOnly GPT-5.4 Nano measured
LMArena Text: coding14041424Qwen3 235B A22B Thinking 2507
LMArena Text: software and IT14021431Qwen3 235B A22B Thinking 2507
Neither model has a result on the other 31 boards here
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index10
Intelligence benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
AA Intelligence Index20.712.7GPT-5.4 Nano
Epoch Capabilities Index146.0Not measuredOnly GPT-5.4 Nano measured
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
Agentic work benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
AA Agentic Index16.01.3GPT-5.4 Nano
Finance Agent (Vals AI)38.2%Not measuredOnly GPT-5.4 Nano measured
Tax Agent Bench (Vals AI)5.0%Not measuredOnly GPT-5.4 Nano measured
Legal Research Bench (Vals AI)6.3%Not measuredOnly GPT-5.4 Nano measured
Excel Modeling Benchmark (Vals AI)44.8%Not measuredOnly GPT-5.4 Nano measured
τ²-Bench Telecom76.0%Not measuredOnly GPT-5.4 Nano measured
Vending-Bench 2Not measured-$11Only Qwen3 235B A22B Thinking 2507 measured
Neither model has a result on the other 35 boards here
ReasoningLike-for-likePublic boards, shown as published, kept out of the index04
Reasoning benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
GPQA Diamond77.5%Not measuredOnly GPT-5.4 Nano measured
MMMU Pro (Vals AI)73.6%Not measuredOnly GPT-5.4 Nano measured
GPQA Diamond (Artificial Analysis run)81.7%Not measuredOnly GPT-5.4 Nano measured
Humanity's Last Exam (Artificial Analysis run)28.3%Not measuredOnly GPT-5.4 Nano measured
GPQA Diamond (Epoch AI run)78.5%80.1%Qwen3 235B A22B Thinking 2507
ARC-AGI51.5%Not measuredOnly GPT-5.4 Nano measured
ARC-AGI-25.7%Not measuredOnly GPT-5.4 Nano measured
Chess puzzles45.0%Not measuredOnly GPT-5.4 Nano measured
Mystery game puzzles9.0%9.0%Tie
LMArena Text: hard prompts13811416Qwen3 235B A22B Thinking 2507
LMArena Text: expert13961463Qwen3 235B A22B Thinking 2507
LMArena Text: hard prompts in English13871432Qwen3 235B A22B Thinking 2507
LMArena Vision: diagrams1210Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: homework1225Not measuredOnly GPT-5.4 Nano measured
Kagi LLM Benchmark39.7%Not measuredOnly GPT-5.4 Nano measured
Neither model has a result on the other 8 boards here
MathLike-for-likePublic boards, shown as published, kept out of the index12
Math benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
AIME88.8%Not measuredOnly GPT-5.4 Nano measured
OTIS Mock AIME87.8%86.7%GPT-5.4 Nano
FrontierMath, tiers 1 to 344.9%Not measuredOnly GPT-5.4 Nano measured
FrontierMath, tier 412.2%Not measuredOnly GPT-5.4 Nano measured
ProofBench5.0%Not measuredOnly GPT-5.4 Nano measured
LMArena Text: math14051412Qwen3 235B A22B Thinking 2507
LMArena Text: mathematical work14091435Qwen3 235B A22B Thinking 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 GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
MMLU Pro77.2%Not measuredOnly GPT-5.4 Nano measured
LegalBench (Vals AI)77.9%Not measuredOnly GPT-5.4 Nano measured
CaseLaw (Vals AI)51.9%Not measuredOnly GPT-5.4 Nano measured
CorpFin (Vals AI)61.2%Not measuredOnly GPT-5.4 Nano measured
TaxEval (Vals AI)67.4%Not measuredOnly GPT-5.4 Nano measured
MortgageTax (Vals AI)59.1%Not measuredOnly GPT-5.4 Nano measured
MedCode (Vals AI)41.0%Not measuredOnly GPT-5.4 Nano measured
SimpleQA Verified11.7%40.4%Qwen3 235B A22B Thinking 2507
LMArena Text: science13741433Qwen3 235B A22B Thinking 2507
LMArena Text: medicine and healthcare13691456Qwen3 235B A22B Thinking 2507
LMArena Text: legal and government13681415Qwen3 235B A22B Thinking 2507
LMArena Text: business and finance13661413Qwen3 235B A22B Thinking 2507
Neither model has a result on the other 8 boards here
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
Web design benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
Design Arena: websitesNot measured1052Only Qwen3 235B A22B Thinking 2507 measured
Design Arena: UI componentsNot measured942Only Qwen3 235B A22B Thinking 2507 measured
Design Arena: game devNot measured956Only Qwen3 235B A22B Thinking 2507 measured
Design Arena: data vizNot measured950Only Qwen3 235B A22B Thinking 2507 measured
Design Arena: 3DNot measured997Only Qwen3 235B A22B Thinking 2507 measured
Design Arena: all code categoriesNot measured1032Only Qwen3 235B A22B Thinking 2507 measured
Neither model has a result on the other 3 boards here
GeneralLike-for-likePublic boards, shown as published, kept out of the index116
General benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 2507Leader
SAGE (Vals AI)38.1%Not measuredOnly GPT-5.4 Nano measured
MedScribe (Vals AI)77.1%Not measuredOnly GPT-5.4 Nano measured
Vals Multimodal Index47.6%Not measuredOnly GPT-5.4 Nano measured
IFBench75.9%Not measuredOnly GPT-5.4 Nano measured
Long context reasoning (AA-LCR)76.7%Not measuredOnly GPT-5.4 Nano measured
DTBench80.3%80.3%Tie
LMCA36.9%29.3%GPT-5.4 Nano
Fiction.LiveBenchNot measured75.0%Only Qwen3 235B A22B Thinking 2507 measured
Creative writing (Lech Mazur)Not measured82.4%Only Qwen3 235B A22B Thinking 2507 measured
LMArena Text13721415Qwen3 235B A22B Thinking 2507
LMArena Text: instruction following13621386Qwen3 235B A22B Thinking 2507
LMArena Text: creative writing13151384Qwen3 235B A22B Thinking 2507
LMArena Vision1196Not measuredOnly GPT-5.4 Nano measured
LMArena Text: English13811429Qwen3 235B A22B Thinking 2507
LMArena Text: non-English13591397Qwen3 235B A22B Thinking 2507
LMArena Text: Chinese13911481Qwen3 235B A22B Thinking 2507
LMArena Text: German13681408Qwen3 235B A22B Thinking 2507
LMArena Text: Japanese13431386Qwen3 235B A22B Thinking 2507
LMArena Text: Korean13211375Qwen3 235B A22B Thinking 2507
LMArena Text: French1395Not measuredOnly GPT-5.4 Nano measured
LMArena Text: Spanish13711389Qwen3 235B A22B Thinking 2507
LMArena Text: Russian13631401Qwen3 235B A22B Thinking 2507
LMArena Text: Polish13661390Qwen3 235B A22B Thinking 2507
LMArena Text: multi-turn13821406Qwen3 235B A22B Thinking 2507
LMArena Text: longer queries13651400Qwen3 235B A22B Thinking 2507
LMArena Text: writing and language13381388Qwen3 235B A22B Thinking 2507
LMArena Text: entertainment and media13251370Qwen3 235B A22B Thinking 2507
LMArena Vision: English1202Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: Chinese1232Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: reading text1204Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: humor1161Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: recognizing things1141Not measuredOnly GPT-5.4 Nano measured
LMArena Vision: creative writing1159Not measuredOnly GPT-5.4 Nano measured
Neither model has a result on the other 15 boards here
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
Image generation benchmarks for GPT-5.4 Nano and Qwen3 235B A22B Thinking 2507
BenchmarkGPT-5.4 NanoQwen3 235B A22B Thinking 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

GPT-5.4 Nano$0.0008Fits in one request
Qwen3 235B A22B Thinking 2507$0.0022Fits in one request

GPT-5.4 Nano costs less

Repository review

50K input + 3K output tokens

GPT-5.4 Nano$0.014Fits in one request
Qwen3 235B A22B Thinking 2507$0.033Fits in one request

GPT-5.4 Nano costs less

Agent loop

200K input + 10K output tokens

GPT-5.4 Nano$0.052Fits in one request
Qwen3 235B A22B Thinking 2507$0.125Exceeds the context window

GPT-5.4 Nano 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-5.4 Nano and Qwen3 235B A22B Thinking 2507
SpecificationOpenAI GPT-5.4 NanoQwen3 235B A22B Thinking 2507
MakerOpenAIQwen
WeightsNo public weights listedOpen weights
ReleasedMarch 17, 2026July 25, 2025
Context window400K128K
Input $/Mtok$0.20$0.45
Output $/Mtok$1.25$3.50
Coding index44.8Not benchmarked
Independent measurements behind it20
Cost per solved task$0.14Not computable
SourceOpenRouter listingOpenRouter listing

Questions

Is GPT-5.4 Nano better than Qwen3 235B A22B Thinking 2507 for coding?

Only GPT-5.4 Nano has a coding score, 44.8 on the Vibe Coding Index; Qwen3 235B A22B Thinking 2507 has none, so the two cannot be ranked on coding. Pick Qwen3 235B A22B Thinking 2507 for work measured by LMArena Text: coding or LMArena Text: software and IT, where it leads.

Which is cheaper, GPT-5.4 Nano or Qwen3 235B A22B Thinking 2507?

GPT-5.4 Nano costs $0.20 per million input tokens and $1.25 per million output tokens; Qwen3 235B A22B Thinking 2507 costs $0.45 and $3.50. Cost per solved task is known for GPT-5.4 Nano only.

Which has the bigger context window, GPT-5.4 Nano or Qwen3 235B A22B Thinking 2507?

GPT-5.4 Nano, with 400K tokens against 128K.

How many benchmarks is the GPT-5.4 Nano vs Qwen3 235B A22B Thinking 2507 comparison based on?

37 benchmarks have a result for both models as of October 22, 2026. GPT-5.4 Nano's coding index rests on 2 independent benchmarks and Qwen3 235B A22B Thinking 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.

Compare next