Phi 4 vs Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 leads Phi 4 on the Artificial Analysis Intelligence Index, 12.0 to 5.9, and Phi 4 costs 5.4x less per output token ($0.14 vs $0.75 per million). Updated October 21, 2026; 12 benchmarks have a result for both.
Phi 4 or Qwen3 235B A22B 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 5.4x cheaper ($0.14 vs $0.75 per million) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.
- Shared results
- 12
- Phi 4 only
- 34
- Qwen3 235B A22B Instruct 2507 only
- 12
- 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.
Coding
Vibe Coding Index, six public coding benchmarks
No call
Neither has a coding score yet.
One side not measured
Intelligence
AA Intelligence Index
Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 leads, 12.0 against 5.9.
Same board, both measured
Web design
Design Arena: websites
No call
Only Qwen3 235B A22B Instruct 2507 has a Design Arena: websites result, 1058.
Directional only
Long documents
Listed context window
Qwen3 235B A22B Instruct 2507
Qwen3 235B A22B Instruct 2507 takes 262K tokens against 16K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Phi 4
Phi 4 $0.0001, Qwen3 235B A22B Instruct 2507 $0.0005 per turn.
List prices
At a glance
- Coding index
- Phi 4 not benchmarked, Qwen3 235B A22B Instruct 2507 not benchmarked.
- Widest category
- Math, 73.0 points apart on average over 2 shared benchmarks.
- Widest single gap
- AA Math Index: Phi 4 18.0 against Qwen3 235B A22B Instruct 2507 91.0, 73.0 points apart on a 0 to 100 scale.
- Output price
- Phi 4 $0.14 against Qwen3 235B A22B Instruct 2507 $0.75 per million output tokens, 5.4x apart.
- Context window
- Phi 4 takes 16K tokens, Qwen3 235B A22B 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.
| Category | Phi 4 | Bars | Qwen3 235B A22B Instruct 2507 | Ties |
|---|---|---|---|---|
| Coding | 0 | 2 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 0 | 2 | ||
| Reasoning | 0 | 2 | ||
| Math | 0 | 2 | ||
| Knowledge | 0 | 1 | ||
| General | 0 | 2 | ||
| All shared boards | 0 | 12 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index02
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| LiveCodeBench (Artificial Analysis run) | 23.1% | 78.8% | Qwen3 235B A22B Instruct 2507 |
| SciCode | Not measured | 41.4% | Only Qwen3 235B A22B Instruct 2507 measured |
| Terminal-Bench Hard (Artificial Analysis run) | 3.8% | 15.2% | Qwen3 235B A22B Instruct 2507 |
| WeirdML | Not measured | 38.7% | Only Qwen3 235B A22B Instruct 2507 measured |
| LMArena Text: coding | 1231 | Not measured | Only Phi 4 measured |
| LMArena Text: software and IT | 1228 | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 33 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AA Intelligence Index | 5.9 | 12.0 | Qwen3 235B A22B Instruct 2507 |
| Epoch Capabilities Index | 130.5 | Not measured | Only Phi 4 measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | 0.0% | 53.2% | Qwen3 235B A22B Instruct 2507 |
| BALROG | 11.6% | Not measured | Only Phi 4 measured |
| Berkeley Function Calling Leaderboard (BFCL) | 28.8% | 52.1% | Qwen3 235B A22B Instruct 2507 |
| Neither model has a result on the other 38 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | 57.5% | 79.0% | Qwen3 235B A22B Instruct 2507 |
| Humanity's Last Exam (Artificial Analysis run) | 3.8% | 15.9% | Qwen3 235B A22B Instruct 2507 |
| GPQA Diamond (Epoch AI run) | 56.1% | Not measured | Only Phi 4 measured |
| ARC-AGI | Not measured | 11.0% | Only Qwen3 235B A22B Instruct 2507 measured |
| ARC-AGI-2 | Not measured | 1.3% | Only Qwen3 235B A22B Instruct 2507 measured |
| LMArena Text: hard prompts | 1220 | Not measured | Only Phi 4 measured |
| LMArena Text: expert | 1203 | Not measured | Only Phi 4 measured |
| LMArena Text: hard prompts in English | 1230 | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 14 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| AA Math Index | 18.0 | 91.0 | Qwen3 235B A22B Instruct 2507 |
| AIME 2025 (Artificial Analysis run) | 18.0% | 91.0% | Qwen3 235B A22B Instruct 2507 |
| OTIS Mock AIME | 13.8% | Not measured | Only Phi 4 measured |
| MATH level 5 | 64.9% | Not measured | Only Phi 4 measured |
| LMArena Text: math | 1246 | Not measured | Only Phi 4 measured |
| LMArena Text: mathematical work | 1254 | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | 71.4% | 84.3% | Qwen3 235B A22B Instruct 2507 |
| MMLU | 84.8% | Not measured | Only Phi 4 measured |
| LMArena Text: science | 1206 | Not measured | Only Phi 4 measured |
| LMArena Text: medicine and healthcare | 1202 | Not measured | Only Phi 4 measured |
| LMArena Text: legal and government | 1241 | Not measured | Only Phi 4 measured |
| LMArena Text: business and finance | 1199 | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 12 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1058 | Only Qwen3 235B A22B Instruct 2507 measured |
| Design Arena: UI components | Not measured | 965 | Only Qwen3 235B A22B Instruct 2507 measured |
| Design Arena: game dev | Not measured | 952 | Only Qwen3 235B A22B Instruct 2507 measured |
| Design Arena: data viz | Not measured | 1069 | Only Qwen3 235B A22B Instruct 2507 measured |
| Design Arena: 3D | Not measured | 994 | Only Qwen3 235B A22B Instruct 2507 measured |
| Design Arena: all code categories | Not measured | 1038 | Only Qwen3 235B A22B Instruct 2507 measured |
| Neither model has a result on the other 3 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index02
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| IFBench | 23.5% | 51.2% | Qwen3 235B A22B Instruct 2507 |
| Long context reasoning (AA-LCR) | 0.0% | 72.0% | Qwen3 235B A22B Instruct 2507 |
| DTBench | Not measured | 78.4% | Only Qwen3 235B A22B Instruct 2507 measured |
| LMCA | Not measured | 23.5% | Only Qwen3 235B A22B Instruct 2507 measured |
| Creative writing (Lech Mazur) | 62.6% | Not measured | Only Phi 4 measured |
| LMArena Text | 1217 | Not measured | Only Phi 4 measured |
| LMArena Text: instruction following | 1201 | Not measured | Only Phi 4 measured |
| LMArena Text: creative writing | 1181 | Not measured | Only Phi 4 measured |
| LMArena Text: English | 1231 | Not measured | Only Phi 4 measured |
| LMArena Text: non-English | 1197 | Not measured | Only Phi 4 measured |
| LMArena Text: Chinese | 1212 | Not measured | Only Phi 4 measured |
| LMArena Text: German | 1222 | Not measured | Only Phi 4 measured |
| LMArena Text: Japanese | 1158 | Not measured | Only Phi 4 measured |
| LMArena Text: Korean | 1151 | Not measured | Only Phi 4 measured |
| LMArena Text: French | 1224 | Not measured | Only Phi 4 measured |
| LMArena Text: Spanish | 1234 | Not measured | Only Phi 4 measured |
| LMArena Text: Russian | 1209 | Not measured | Only Phi 4 measured |
| LMArena Text: multi-turn | 1206 | Not measured | Only Phi 4 measured |
| LMArena Text: longer queries | 1217 | Not measured | Only Phi 4 measured |
| LMArena Text: writing and language | 1198 | Not measured | Only Phi 4 measured |
| LMArena Text: entertainment and media | 1162 | Not measured | Only Phi 4 measured |
| Neither model has a result on the other 25 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Phi 4 | Qwen3 235B A22B Instruct 2507 | Leader |
|---|---|---|---|
| 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 costs less
Repository review
50K input + 3K output tokens
Phi 4 costs less
Agent loop
200K input + 10K output tokens
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.
| Specification | Microsoft Phi 4 | Qwen3 235B A22B Instruct 2507 |
|---|---|---|
| Maker | Microsoft | Qwen |
| Weights | Open weights | Open weights |
| Released | January 10, 2025 | July 21, 2025 |
| Context window | 16K | 262K |
| Input $/Mtok | $0.07 | $0.15 |
| Output $/Mtok | $0.14 | $0.75 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 0 | 0 |
| Cost per solved task | Not computable | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Phi 4 better than Qwen3 235B A22B 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 5.4x cheaper ($0.14 vs $0.75 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 235B A22B Instruct 2507?
Phi 4 costs $0.07 per million input tokens and $0.14 per million output tokens; Qwen3 235B A22B Instruct 2507 costs $0.15 and $0.75. Cost per solved task is known for neither of them.
Which has the bigger context window, Phi 4 or Qwen3 235B A22B Instruct 2507?
Qwen3 235B A22B Instruct 2507, with 262K tokens against 16K.
How many benchmarks is the Phi 4 vs Qwen3 235B A22B Instruct 2507 comparison based on?
12 benchmarks have a result for both models as of October 21, 2026. Phi 4's coding index rests on 0 independent benchmarks and Qwen3 235B A22B Instruct 2507's on 0. Where only one model has a result, the page says so instead of naming a winner.