# Command A vs GPT-4

> Command A is the stronger model overall, scoring 35.9 to 28.9 on the Noometry Index.

- Canonical page: https://noometry.com/compare/command-a-vs-gpt-4
- Last updated: 2026-10-10
- Shared benchmarks: 20

## Summary

- They share 20 benchmarks with published results for both. Command A scores higher in 7 categories and GPT-4 in 1 category; 7 gaps are clear of the uncertainty.
- The widest gap is in math, where Command A leads 33.5 to 10.1.
- The biggest single-benchmark swing is LMCA: 10.3% for Command A and 17.1% for GPT-4.
- Command A is cheaper at $2.50 / $10 per million input/output tokens, against $30 / $60 for GPT-4.
- Command A accepts more context: 256K tokens versus 8K.
- Command A has downloadable open weights; the other is API-only.

## Snapshot

| | Command A | GPT-4 |
|---|---|---|
| Provider | Cohere | OpenAI |
| Noometry Index | 35.9 | 28.9 |
| Rank | 212 | 308 |
| Context | 256K | 8K |
| Input $/M | $2.50 | $30 |
| Output $/M | $10 | $60 |
| Weights | Open | Proprietary |

## Coding

- Command A: 26.8 (#317)
- GPT-4: 31.2 (#277)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Coding | 1331 | 1254 |
| Aider Polyglot | 12% | — |
| WeirdML | — | 12.4% |
| BigCodeBench Instruct | — | 46% |
| BigCodeBench Complete | — | 57.2% |
| HumanEval+ | — | 79.3% |

## Agentic & Tool Use

- Command A: 35.7 (#36)
- GPT-4: —

| Benchmark | Command A | GPT-4 |
|---|---|---|
| Berkeley Function Calling Leaderboard | 57.1% | — |
| METR Time Horizons | — | 36.1% |

## Reasoning

- Command A: 17.7 (#276)
- GPT-4: 17.3 (#285)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Hard Prompts | 1326 | 1241 |
| DTBench | 61.3% | 62.7% |
| LMCA | 10.3% | 17.1% |
| Kagi LLM Benchmark | 28.8% | — |
| Chess Puzzles | — | 4% |
| Mystery Game Puzzles | — | 12% |
| BIG-Bench Hard | — | 75.1% |
| Epoch Capabilities Index | — | 125.88 |
| ForecastBench | — | 57.8 |
| HellaSwag | — | 95.3% |
| WinoGrande | — | 87.5% |

## Math

- Command A: 33.5 (#172)
- GPT-4: 10.1 (#304)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Math | 1300 | 1269 |
| OTIS Mock AIME 2024-2025 | — | 1.1% |
| MATH Level 5 | — | 23% |
| GSM8K | — | 92% |

## Knowledge

- Command A: 36.5 (#157)
- GPT-4: 18.4 (#277)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Expert | 1295 | 1211 |
| GPQA Diamond | — | 35.7% |
| Vectara Hallucination Rate | 9.3% | — |
| MMLU | — | 86.4% |
| TriviaQA | — | 84.8% |

## Multilingual

- Command A: 45.5 (#166)
- GPT-4: 40.8 (#210)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Non-English | 1313 | 1246 |
| LMArena Chinese | 1327 | 1242 |
| LMArena French | 1352 | 1282 |
| LMArena German | 1341 | 1251 |
| LMArena Japanese | 1283 | 1209 |
| LMArena Korean | 1285 | 1184 |
| LMArena Russian | 1314 | 1251 |
| LMArena Spanish | 1346 | 1261 |

## Instruction Following

- Command A: 69.3 (#172)
- GPT-4: 65.5 (#217)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Instruction Following | 1309 | 1241 |

## Long Context

- Command A: 40.6 (#147)
- GPT-4: 37.7 (#207)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Longer Query | 1334 | 1244 |

## Writing & Preference

- Command A: 47.8 (#204)
- GPT-4: 35.0 (#263)

| Benchmark | Command A | GPT-4 |
|---|---|---|
| LMArena Text | 1331 | 1263 |
| LMArena Creative Writing | 1319 | 1244 |
| EQ-Bench Creative Writing | 1145 | 752 |
| LMArena Multi-Turn | 1338 | 1257 |

## FAQ

### Is Command A better than GPT-4?

Command A is the stronger model overall, scoring 35.9 to 28.9 on the Noometry Index.

### Which is cheaper, Command A or GPT-4?

Command A is cheaper. It lists at $2.50 per million input tokens and $10 per million output tokens; GPT-4 lists at $30 and $60.

### Is Command A or GPT-4 better for coding?

GPT-4 scores higher on coding benchmarks: 31.2 versus 26.8 in the Noometry coding category.

### Which has the bigger context window?

Command A does, with 256K tokens against 8K.

### How many benchmarks do Command A and GPT-4 share?

20 benchmarks have published results for both models. Command A has 24 scored results on Noometry and GPT-4 has 38.