Command A vs DeepSeek V3.2 Exp
DeepSeek V3.2 Exp leads Command A on the Artificial Analysis Intelligence Index, 16.6 to 13.1, and DeepSeek V3.2 Exp costs 24.4x less per output token ($0.41 vs $10.00 per million). Updated October 21, 2026; 41 benchmarks have a result for both.
Command A or DeepSeek V3.2 Exp: which to pick
Neither model has a coding score yet, so the two cannot be ranked on coding.
Pick Command A for a larger context window (256K against 164K tokens) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.
- Shared results
- 41
- Command A only
- 19
- DeepSeek V3.2 Exp only
- 16
- 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
DeepSeek V3.2 Exp
DeepSeek V3.2 Exp leads, 16.6 against 13.1.
Same board, both measured
Agentic work
AA Agentic Index
No call
Only Command A has a AA Agentic Index result, 1.0.
Directional only
Reasoning
GPQA Diamond
No call
Only Command A has a GPQA Diamond result, 48.5%.
Directional only
Math
AIME
No call
Only Command A has a AIME result, 13.3%.
Directional only
Knowledge
MMLU Pro
No call
Only Command A has a MMLU Pro result, 69.2%.
Directional only
Web design
Design Arena: websites
No call
Only DeepSeek V3.2 Exp has a Design Arena: websites result, 1179.
Directional only
Long documents
Listed context window
Command A
Command A takes 256K tokens against 164K.
Listed by OpenRouter
Everyday cost
1K input + 500 output tokens
DeepSeek V3.2 Exp
Command A $0.0075, DeepSeek V3.2 Exp $0.0005 per turn.
List prices
Cost per solved task
One SWE-bench Verified task, solved
No call
Only Command A has the number, $2.69.
Directional only
At a glance
- Coding index
- Command A not benchmarked, DeepSeek V3.2 Exp not benchmarked.
- Widest category
- Math, 47.5 points apart on average over 4 shared benchmarks.
- Widest single gap
- AA Math Index: Command A 13.0 against DeepSeek V3.2 Exp 87.7, 74.7 points apart on a 0 to 100 scale.
- Output price
- Command A $10.00 against DeepSeek V3.2 Exp $0.41 per million output tokens, 24.4x apart.
- Context window
- Command A takes 256K tokens, DeepSeek V3.2 Exp 164K.
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 | Command A | Bars | DeepSeek V3.2 Exp | Ties |
|---|---|---|---|---|
| Coding | 0 | 4 | ||
| Intelligence | 0 | 1 | ||
| Agentic work | 1 | 0 | ||
| Reasoning | 0 | 5 | ||
| Math | 0 | 4 | ||
| Knowledge | 0 | 5 | ||
| General | 1 | 20 | ||
| All shared boards | 2 | 39 |
CodingLike-for-likeSix of these boards feed the Vibe Coding Index04
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| Aider polyglot | Not measured | 74.2% | Only DeepSeek V3.2 Exp measured |
| LMArena WebDev | Not measured | 1272 | Only DeepSeek V3.2 Exp measured |
| AA Coding Index | 27.8 | Not measured | Only Command A measured |
| LiveCodeBench (Vals AI) | 35.1% | Not measured | Only Command A measured |
| SWE-bench Verified (Vals AI) | 7.8% | Not measured | Only Command A measured |
| Terminal-Bench 2.1 (Vals AI) | 17.6% | Not measured | Only Command A measured |
| LiveCodeBench (Artificial Analysis run) | 28.7% | 78.9% | DeepSeek V3.2 Exp |
| SciCode | 38.5% | Not measured | Only Command A measured |
| Terminal-Bench Hard (Artificial Analysis run) | 25.0% | 31.1% | DeepSeek V3.2 Exp |
| WeirdML | Not measured | 39.5% | Only DeepSeek V3.2 Exp measured |
| LMArena Text: coding | 1330 | 1438 | DeepSeek V3.2 Exp |
| LMArena Text: software and IT | 1336 | 1438 | DeepSeek V3.2 Exp |
| LMArena WebDev: HTML | Not measured | 1281 | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 25 boards here | |||
IntelligenceLike-for-likePublic boards, shown as published, kept out of the index01
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| AA Intelligence Index | 13.1 | 16.6 | DeepSeek V3.2 Exp |
| Epoch Capabilities Index | Not measured | 144.9 | Only DeepSeek V3.2 Exp measured |
Agentic workLike-for-likePublic boards, shown as published, kept out of the index10
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| AA Agentic Index | 1.0 | Not measured | Only Command A measured |
| Finance Agent (Vals AI) | 9.0% | Not measured | Only Command A measured |
| τ²-Bench Telecom | 80.7% | 33.9% | Command A |
| TheAgentCompany | Not measured | 42.9% | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 35 boards here | |||
ReasoningLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| GPQA Diamond | 48.5% | Not measured | Only Command A measured |
| GPQA Diamond (Artificial Analysis run) | 76.1% | 79.7% | DeepSeek V3.2 Exp |
| Humanity's Last Exam (Artificial Analysis run) | 12.0% | 14.9% | DeepSeek V3.2 Exp |
| LMArena Text: hard prompts | 1326 | 1429 | DeepSeek V3.2 Exp |
| LMArena Text: expert | 1296 | 1422 | DeepSeek V3.2 Exp |
| LMArena Text: hard prompts in English | 1335 | 1444 | DeepSeek V3.2 Exp |
| Neither model has a result on the other 15 boards here | |||
MathLike-for-likePublic boards, shown as published, kept out of the index04
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| AIME | 13.3% | Not measured | Only Command A measured |
| MATH 500 (Vals AI) | 76.2% | Not measured | Only Command A measured |
| MGSM (Vals AI) | 85.7% | Not measured | Only Command A measured |
| AA Math Index | 13.0 | 87.7 | DeepSeek V3.2 Exp |
| AIME 2025 (Artificial Analysis run) | 13.0% | 87.7% | DeepSeek V3.2 Exp |
| LMArena Text: math | 1300 | 1422 | DeepSeek V3.2 Exp |
| LMArena Text: mathematical work | 1312 | 1433 | DeepSeek V3.2 Exp |
| Neither model has a result on the other 6 boards here | |||
KnowledgeLike-for-likePublic boards, shown as published, kept out of the index05
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| MMLU Pro | 69.2% | Not measured | Only Command A measured |
| LegalBench (Vals AI) | 79.7% | Not measured | Only Command A measured |
| CaseLaw (Vals AI) | 64.5% | Not measured | Only Command A measured |
| CorpFin (Vals AI) | 46.0% | Not measured | Only Command A measured |
| TaxEval (Vals AI) | 61.4% | Not measured | Only Command A measured |
| MedQA (Vals AI) | 80.5% | Not measured | Only Command A measured |
| MedCode (Vals AI) | 19.7% | Not measured | Only Command A measured |
| MMLU Pro (Artificial Analysis run) | 71.2% | 85.0% | DeepSeek V3.2 Exp |
| LMArena Text: science | 1339 | 1443 | DeepSeek V3.2 Exp |
| LMArena Text: medicine and healthcare | 1334 | 1468 | DeepSeek V3.2 Exp |
| LMArena Text: legal and government | 1348 | 1419 | DeepSeek V3.2 Exp |
| LMArena Text: business and finance | 1315 | 1412 | DeepSeek V3.2 Exp |
| Neither model has a result on the other 5 boards here | |||
Web designDirectional onlyPublic boards, shown as published, kept out of the index7 boards, none shared
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| Design Arena: websites | Not measured | 1179 | Only DeepSeek V3.2 Exp measured |
| Design Arena: UI components | Not measured | 1167 | Only DeepSeek V3.2 Exp measured |
| Design Arena: game dev | Not measured | 1144 | Only DeepSeek V3.2 Exp measured |
| Design Arena: data viz | Not measured | 1152 | Only DeepSeek V3.2 Exp measured |
| Design Arena: 3D | Not measured | 1148 | Only DeepSeek V3.2 Exp measured |
| Design Arena: all code categories | Not measured | 1171 | Only DeepSeek V3.2 Exp measured |
| Design Arena: SVG | Not measured | 1032 | Only DeepSeek V3.2 Exp measured |
| Neither model has a result on the other 2 boards here | |||
GeneralLike-for-likePublic boards, shown as published, kept out of the index120
| Benchmark | Command A | DeepSeek V3.2 Exp | Leader |
|---|---|---|---|
| MedScribe (Vals AI) | 55.7% | Not measured | Only Command A measured |
| IFBench | 73.9% | 54.1% | Command A |
| Long context reasoning (AA-LCR) | 52.7% | 72.3% | DeepSeek V3.2 Exp |
| DTBench | 61.3% | 87.7% | DeepSeek V3.2 Exp |
| LMCA | 10.3% | 29.1% | DeepSeek V3.2 Exp |
| Fiction.LiveBench | Not measured | 83.3% | Only DeepSeek V3.2 Exp measured |
| CL-bench | Not measured | 13.2% | Only DeepSeek V3.2 Exp measured |
| CL-bench Life | Not measured | 9.5% | Only DeepSeek V3.2 Exp measured |
| LMArena Text | 1331 | 1425 | DeepSeek V3.2 Exp |
| LMArena Text: instruction following | 1309 | 1401 | DeepSeek V3.2 Exp |
| LMArena Text: creative writing | 1319 | 1404 | DeepSeek V3.2 Exp |
| LMArena Text: English | 1344 | 1441 | DeepSeek V3.2 Exp |
| LMArena Text: non-English | 1313 | 1404 | DeepSeek V3.2 Exp |
| LMArena Text: Chinese | 1327 | 1461 | DeepSeek V3.2 Exp |
| LMArena Text: German | 1342 | 1440 | DeepSeek V3.2 Exp |
| LMArena Text: Japanese | 1285 | 1331 | DeepSeek V3.2 Exp |
| LMArena Text: Korean | 1285 | 1372 | DeepSeek V3.2 Exp |
| LMArena Text: French | 1351 | 1434 | DeepSeek V3.2 Exp |
| LMArena Text: Spanish | 1347 | 1441 | DeepSeek V3.2 Exp |
| LMArena Text: Russian | 1314 | 1412 | DeepSeek V3.2 Exp |
| LMArena Text: Polish | 1324 | 1435 | DeepSeek V3.2 Exp |
| LMArena Text: multi-turn | 1338 | 1422 | DeepSeek V3.2 Exp |
| LMArena Text: longer queries | 1334 | 1418 | DeepSeek V3.2 Exp |
| LMArena Text: writing and language | 1328 | 1407 | DeepSeek V3.2 Exp |
| LMArena Text: entertainment and media | 1306 | 1407 | DeepSeek V3.2 Exp |
| Neither model has a result on the other 21 boards here | |||
Image generationNot comparablePublic boards, shown as published, kept out of the index0 boards, none shared
| Benchmark | Command A | DeepSeek V3.2 Exp | 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
DeepSeek V3.2 Exp costs less
Repository review
50K input + 3K output tokens
DeepSeek V3.2 Exp costs less
Agent loop
200K input + 10K output tokens
DeepSeek V3.2 Exp 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 | Cohere Command A | DeepSeek V3.2 Exp |
|---|---|---|
| Maker | Cohere | DeepSeek |
| Weights | Open weights | Open weights |
| Released | March 13, 2025 | September 29, 2025 |
| Context window | 256K | 164K |
| Input $/Mtok | $2.50 | $0.27 |
| Output $/Mtok | $10.00 | $0.41 |
| Coding index | Not benchmarked | Not benchmarked |
| Independent measurements behind it | 1 | 1 |
| Cost per solved task | $2.69 | Not computable |
| Source | OpenRouter listing | OpenRouter listing |
Questions
Is Command A better than DeepSeek V3.2 Exp for coding?
Neither model has a coding score yet, so the two cannot be ranked on coding. Pick Command A for a larger context window (256K against 164K tokens) and work measured by LiveCodeBench (Artificial Analysis run) or Terminal-Bench Hard (Artificial Analysis run), where it leads.
Which is cheaper, Command A or DeepSeek V3.2 Exp?
Command A costs $2.50 per million input tokens and $10.00 per million output tokens; DeepSeek V3.2 Exp costs $0.27 and $0.41. Cost per solved task is known for Command A only.
Which has the bigger context window, Command A or DeepSeek V3.2 Exp?
Command A, with 256K tokens against 164K.
How many benchmarks is the Command A vs DeepSeek V3.2 Exp comparison based on?
41 benchmarks have a result for both models as of October 21, 2026. Command A's coding index rests on 1 independent benchmark and DeepSeek V3.2 Exp's on 1. Where only one model has a result, the page says so instead of naming a winner.