Model comparison
Amazon Nova Lite vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 31.9 on the Noometry Index.
Last verified . 17 shared benchmarks.
DeepSeekSummary
- They share 17 benchmarks with published results for both. Amazon Nova Lite scores higher in 1 category and DeepSeek-V2.5 (Sep 2024) in 7 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in knowledge, where DeepSeek-V2.5 (Sep 2024) leads 34.8 to 26.3.
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | Amazon | DeepSeek |
| Noometry Index | 31.9 | 37.6 |
| Released | 2024-12-03 | 2024-09-06 |
| Weights | Proprietary | Open |
| Context window | 300K | — |
| Max output | 10K | — |
| Input $ / M tokens | $0.06 | — |
| Output $ / M tokens | $0.24 | — |
| Results tracked | 34 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
Amazon Nova Lite: 32.5 (#270), DeepSeek-V2.5 (Sep 2024): 31.7 (#281)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Coding | 1239 | 1309 |
| Aider Polyglot | — | 17.8% |
| BigCodeBench Instruct | — | 48.6% |
| LiveBench Coding | 27.5% | — |
| BigCodeBench Complete | — | 53.2% |
| ALE-Bench | 236.25 | — |
| HumanEval+ | — | 83.5% |
| MBPP+ | — | 74.1% |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 19.3 (#260), DeepSeek-V2.5 (Sep 2024): 25.6 (#145)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Hard Prompts | 1220 | 1289 |
| LiveBench Reasoning | 36.7% | — |
| LiveBench Data Analysis | 37.2% | — |
| LiveBench | 36.4% | — |
Math DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 27.8 (#247), DeepSeek-V2.5 (Sep 2024): 35.9 (#177)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | 1227 | 1288 |
| Omni-MATH | 23.3% | — |
| LiveBench Math | 36.7% | — |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 26.3 (#259), DeepSeek-V2.5 (Sep 2024): 34.8 (#193)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | 1201 | 1266 |
| Humanity's Last Exam | 3.6% | — |
| MMLU-Pro | 60% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 39.7% | — |
| MMLU | 77% | — |
Multimodal Not comparable
Amazon Nova Lite: 25.5 (#123), DeepSeek-V2.5 (Sep 2024): —
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Vision | 990 | — |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 38.0 (#232), DeepSeek-V2.5 (Sep 2024): 42.5 (#193)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | 1208 | 1273 |
| LMArena Chinese | 1225 | 1318 |
| LMArena French | 1237 | 1289 |
| LMArena German | 1229 | 1258 |
| LMArena Japanese | 1153 | 1228 |
| LMArena Korean | 1153 | 1209 |
| LMArena Russian | 1216 | 1289 |
| LMArena Spanish | 1226 | 1248 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 59.3 (#255), DeepSeek-V2.5 (Sep 2024): 67.5 (#194)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | 1205 | 1280 |
| LiveBench Instruction Following | 54.1% | — |
| IFEval | 77.6% | — |
Long Context DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 37.4 (#217), DeepSeek-V2.5 (Sep 2024): 39.5 (#174)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | 1234 | 1301 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
Amazon Nova Lite: 42.1 (#237), DeepSeek-V2.5 (Sep 2024): 49.8 (#187)
| Benchmark | Amazon Nova Lite | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | 1229 | 1294 |
| LMArena Creative Writing | 1197 | 1285 |
| LMArena Multi-Turn | 1199 | 1297 |
| WildBench | 75% | — |
| LiveBench Language | 25.9% | — |
Frequently asked questions
Is Amazon Nova Lite better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.6 to 31.9 on the Noometry Index.
Is Amazon Nova Lite or DeepSeek-V2.5 (Sep 2024) better for coding?
They score almost the same on coding (32.5 vs 31.7); test both on your own repository before choosing.
How many benchmarks do Amazon Nova Lite and DeepSeek-V2.5 (Sep 2024) share?
17 benchmarks have published results for both models. Amazon Nova Lite has 34 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.