Model comparison
Command R+ vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 32.4 on the Noometry Index.
Last verified . 21 shared benchmarks.
Cohere
DeepSeekSummary
- They share 21 benchmarks with published results for both. Command R+ scores higher in 1 category and DeepSeek-V2.5 (Sep 2024) in 7 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where DeepSeek-V2.5 (Sep 2024) leads 25.6 to 9.2.
- The biggest single-benchmark swing is BigCodeBench Instruct: 33.8% for Command R+ and 48.6% for DeepSeek-V2.5 (Sep 2024).
Side by side
| Command R+ | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | Cohere | DeepSeek |
| Noometry Index | 32.4 | 37.6 |
| Released | 2024-08-30 | 2024-09-06 |
| Weights | Open | Open |
| Context window | 128K | — |
| Max output | 4K | — |
| Input $ / M tokens | $2.50 | — |
| Output $ / M tokens | $10 | — |
| Results tracked | 34 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2.5 (Sep 2024) leads
Command R+: 29.1 (#309), DeepSeek-V2.5 (Sep 2024): 31.7 (#281)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| BigCodeBench Instruct | 33.8% | 48.6% |
| LMArena Coding | 1187 | 1309 |
| BigCodeBench Complete | 41.9% | 53.2% |
| HumanEval+ | 56.7% | 83.5% |
| MBPP+ | 63.5% | 74.1% |
| Aider Polyglot | — | 17.8% |
| LiveBench Coding | 19.1% | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Command R+: 9.2 (#344), DeepSeek-V2.5 (Sep 2024): 25.6 (#145)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Hard Prompts | 1186 | 1289 |
| SimpleBench | 17.4% | — |
| LiveBench Reasoning | 24.8% | — |
| DTBench | 54.9% | — |
| LiveBench Data Analysis | 38.1% | — |
| LMCA | 5% | — |
| Epoch Capabilities Index | 119.34 | — |
| LiveBench | 31.8% | — |
Math DeepSeek-V2.5 (Sep 2024) leads
Command R+: 28.9 (#242), DeepSeek-V2.5 (Sep 2024): 35.9 (#177)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | 1188 | 1288 |
| LiveBench Math | 21.3% | — |
Knowledge Command R+ leads
Command R+: 36.4 (#169), DeepSeek-V2.5 (Sep 2024): 34.8 (#193)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | 1174 | 1266 |
| Vectara Hallucination Rate | 6.9% | — |
| MMLU | 69.4% | — |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
Command R+: 38.6 (#227), DeepSeek-V2.5 (Sep 2024): 42.5 (#193)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | 1216 | 1273 |
| LMArena Chinese | 1226 | 1318 |
| LMArena French | 1209 | 1289 |
| LMArena German | 1216 | 1258 |
| LMArena Japanese | 1166 | 1228 |
| LMArena Korean | 1138 | 1209 |
| LMArena Russian | 1227 | 1289 |
| LMArena Spanish | 1189 | 1248 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
Command R+: 60.0 (#254), DeepSeek-V2.5 (Sep 2024): 67.5 (#194)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | 1197 | 1280 |
| LiveBench Instruction Following | 57.6% | — |
Long Context DeepSeek-V2.5 (Sep 2024) leads
Command R+: 37.3 (#219), DeepSeek-V2.5 (Sep 2024): 39.5 (#174)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | 1230 | 1301 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
Command R+: 43.5 (#228), DeepSeek-V2.5 (Sep 2024): 49.8 (#187)
| Benchmark | Command R+ | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | 1229 | 1294 |
| LMArena Creative Writing | 1235 | 1285 |
| LMArena Multi-Turn | 1213 | 1297 |
| LiveBench Language | 29.7% | — |
Frequently asked questions
Is Command R+ better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 32.4 on the Noometry Index.
Is Command R+ or DeepSeek-V2.5 (Sep 2024) better for coding?
DeepSeek-V2.5 (Sep 2024) scores higher on coding benchmarks: 31.7 versus 29.1 in the Noometry coding category.
How many benchmarks do Command R+ and DeepSeek-V2.5 (Sep 2024) share?
21 benchmarks have published results for both models. Command R+ has 34 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.