DeepSeek V3 vs Phi 4
Updated October 21, 2026. 2 benchmarks have a result for both. Coding is ranked on the Vibe Coding Index; every other board is shown as published.
Decision reading
Neither model has a coding score yet, so the two cannot be ranked on coding.
DeepSeek V3 has the edge where the context window is larger, 164K against 16K tokens.
- Shared results
- 2
- DeepSeek V3 only
- 11
- Phi 4 only
- 43
- Like-for-like categories
- 2 of 9
Shared results can support a head-to-head reading. Results only one model has describe coverage, not superiority.
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
No call
Only Phi 4 has a AA Intelligence Index result, 5.9.
Directional only
Web design
Design Arena: websites
No call
Only DeepSeek V3 has a Design Arena: websites result, 1120.
Directional only
Long documents
Listed context window
DeepSeek V3
DeepSeek V3 takes 164K tokens against 16K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
Phi 4
DeepSeek V3 $0.0008, Phi 4 $0.0001 per turn.
List prices
At a glance
- Coding index
- DeepSeek V3 not benchmarked, Phi 4 not benchmarked.
- Widest category
- Math, 35.1 points apart on average over 1 shared benchmark.
- Widest single gap
- OTIS Mock AIME: DeepSeek V3 48.9% against Phi 4 13.8%, 35.1 points apart on a 0 to 100 scale.
- Output price
- DeepSeek V3 $0.9594 against Phi 4 $0.14 per million output tokens, 6.9x apart.
- Context window
- DeepSeek V3 takes 164K tokens, Phi 4 16K.
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 | DeepSeek V3 | Bars | Phi 4 | Ties |
|---|---|---|---|---|
| Reasoning | 1 | 0 | ||
| Math | 1 | 0 | ||
| All shared boards | 2 | 0 |
CodingDirectional onlySix of these boards feed the Vibe Coding Index6 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| Aider polyglot | 70.2% | Not measured | Only DeepSeek V3 measured |
| SWE-bench Lite | 30.7% | Not measured | Only DeepSeek V3 measured |
| LiveCodeBench (Artificial Analysis run) | Not measured | 23.1% | Only Phi 4 measured |
| Terminal-Bench Hard (Artificial Analysis run) | Not measured | 3.8% | Only Phi 4 measured |
| LMArena Text: coding | Not measured | 1231 | Only Phi 4 measured |
| LMArena Text: software and IT | Not measured | 1228 | Only Phi 4 measured |
| Neither model has a result on the other 32 boards here | |||
IntelligenceDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| AA Intelligence Index | Not measured | 5.9 | Only Phi 4 measured |
| Epoch Capabilities Index | Not measured | 130.5 | Only Phi 4 measured |
Agentic workDirectional onlyPublic boards, shown as published, kept out of the index2 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| τ²-Bench Telecom | Not measured | 0.0% | Only Phi 4 measured |
| BALROG | Not measured | 11.6% | Only 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 index10
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| GPQA Diamond (Artificial Analysis run) | Not measured | 57.5% | Only Phi 4 measured |
| Humanity's Last Exam (Artificial Analysis run) | Not measured | 3.8% | Only Phi 4 measured |
| GPQA Diamond (Epoch AI run) | 71.2% | 56.1% | DeepSeek V3 |
| LMArena Text: hard prompts | Not measured | 1220 | Only Phi 4 measured |
| LMArena Text: expert | Not measured | 1203 | Only Phi 4 measured |
| LMArena Text: hard prompts in English | Not measured | 1230 | Only Phi 4 measured |
| Neither model has a result on the other 15 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| AA Math Index | Not measured | 18.0 | Only Phi 4 measured |
| AIME 2025 (Artificial Analysis run) | Not measured | 18.0% | Only Phi 4 measured |
| OTIS Mock AIME | 48.9% | 13.8% | DeepSeek V3 |
| MATH level 5 | Not measured | 64.9% | Only Phi 4 measured |
| LMArena Text: math | Not measured | 1246 | Only Phi 4 measured |
| LMArena Text: mathematical work | Not measured | 1254 | Only Phi 4 measured |
| Neither model has a result on the other 7 boards here | |||
KnowledgeDirectional onlyPublic boards, shown as published, kept out of the index6 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| MMLU Pro (Artificial Analysis run) | Not measured | 71.4% | Only Phi 4 measured |
| MMLU | Not measured | 84.8% | Only Phi 4 measured |
| LMArena Text: science | Not measured | 1206 | Only Phi 4 measured |
| LMArena Text: medicine and healthcare | Not measured | 1202 | Only Phi 4 measured |
| LMArena Text: legal and government | Not measured | 1241 | Only Phi 4 measured |
| LMArena Text: business and finance | Not measured | 1199 | Only Phi 4 measured |
| Neither model has a result on the other 11 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| Design Arena: websites | 1120 | Not measured | Only DeepSeek V3 measured |
| Design Arena: UI components | 1093 | Not measured | Only DeepSeek V3 measured |
| Design Arena: game dev | 1053 | Not measured | Only DeepSeek V3 measured |
| Design Arena: data viz | 1089 | Not measured | Only DeepSeek V3 measured |
| Design Arena: 3D | 1087 | Not measured | Only DeepSeek V3 measured |
| Design Arena: all code categories | 1108 | Not measured | Only DeepSeek V3 measured |
| Design Arena: SVG | 977 | Not measured | Only DeepSeek V3 measured |
| Neither model has a result on the other 2 boards here | |||
GeneralDirectional onlyPublic boards, shown as published, kept out of the index21 boards, none shared
| Benchmark | DeepSeek V3 | Phi 4 | Leader |
|---|---|---|---|
| IFBench | Not measured | 23.5% | Only Phi 4 measured |
| Long context reasoning (AA-LCR) | Not measured | 0.0% | Only Phi 4 measured |
| DTBench | 62.7% | Not measured | Only DeepSeek V3 measured |
| LMCA | 15.2% | Not measured | Only DeepSeek V3 measured |
| Creative writing (Lech Mazur) | Not measured | 62.6% | Only Phi 4 measured |
| LMArena Text | Not measured | 1217 | Only Phi 4 measured |
| LMArena Text: instruction following | Not measured | 1201 | Only Phi 4 measured |
| LMArena Text: creative writing | Not measured | 1181 | Only Phi 4 measured |
| LMArena Text: English | Not measured | 1231 | Only Phi 4 measured |
| LMArena Text: non-English | Not measured | 1197 | Only Phi 4 measured |
| LMArena Text: Chinese | Not measured | 1212 | Only Phi 4 measured |
| LMArena Text: German | Not measured | 1222 | Only Phi 4 measured |
| LMArena Text: Japanese | Not measured | 1158 | Only Phi 4 measured |
| LMArena Text: Korean | Not measured | 1151 | Only Phi 4 measured |
| LMArena Text: French | Not measured | 1224 | Only Phi 4 measured |
| LMArena Text: Spanish | Not measured | 1234 | Only Phi 4 measured |
| LMArena Text: Russian | Not measured | 1209 | Only Phi 4 measured |
| LMArena Text: multi-turn | Not measured | 1206 | Only Phi 4 measured |
| LMArena Text: longer queries | Not measured | 1217 | Only Phi 4 measured |
| LMArena Text: writing and language | Not measured | 1198 | Only Phi 4 measured |
| LMArena Text: entertainment and media | Not measured | 1162 | 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 | DeepSeek V3 | Phi 4 | 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 | DeepSeek V3 | Microsoft Phi 4 |
|---|---|---|
| Maker | DeepSeek | Microsoft |
| Weights | Open weights | Open weights |
| Released | December 26, 2024 | January 10, 2025 |
| Context window | 164K | 16K |
| Input $/Mtok | $0.2887 | $0.07 |
| Output $/Mtok | $0.9594 | $0.14 |
| 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 DeepSeek V3 better than Phi 4 for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. DeepSeek V3 has the edge where the context window is larger, 164K against 16K tokens.
Which is cheaper, DeepSeek V3 or Phi 4?
DeepSeek V3 costs $0.2887 per million input tokens and $0.9594 per million output tokens; Phi 4 costs $0.07 and $0.14. Cost per solved task is known for neither of them.
Which has the bigger context window, DeepSeek V3 or Phi 4?
DeepSeek V3, with 164K tokens against 16K.
How many benchmarks is the DeepSeek V3 vs Phi 4 comparison based on?
2 benchmarks have a result for both models as of October 21, 2026. DeepSeek V3's coding index rests on 0 independent benchmarks and Phi 4's on 0. Where only one model has a result, the page says so instead of naming a winner.