Model comparison
Claude 3.5 Haiku vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 28.9 on the Noometry Index.
Last verified . 20 shared benchmarks.
DeepSeekSummary
- They share 20 benchmarks with published results for both. Claude 3.5 Haiku 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 math, where DeepSeek-V2.5 (Sep 2024) leads 33.3 to 13.7.
- The biggest single-benchmark swing is Aider Polyglot: 28% for Claude 3.5 Haiku and 17.8% for DeepSeek-V2.5 (Sep 2024).
- DeepSeek-V2.5 (Sep 2024) has downloadable open weights; the other is API-only.
Side by side
| Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 28.9 | 37.0 |
| Released | 2024-10-22 | 2024-09-06 |
| Weights | Proprietary | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 49 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Claude 3.5 Haiku leads
Claude 3.5 Haiku: 32.6 (#258), DeepSeek-V2.5 (Sep 2024): 31.3 (#274)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| Aider Polyglot | 28% | 17.8% |
| BigCodeBench Instruct | 46.1% | 48.6% |
| LMArena Coding | 1287 | 1310 |
| BigCodeBench Complete | 59% | 53.2% |
| SciCode | 27.4% | — |
| WeirdML | 30.7% | — |
| LiveBench Coding | 51.4% | — |
| CadEval | 32% | — |
| HumanEval+ | — | 83.5% |
| MBPP+ | — | 74.1% |
Agentic & Tool Use Not comparable
Claude 3.5 Haiku: 27.7 (#90), DeepSeek-V2.5 (Sep 2024): —
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| BALROG | 19.3% | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 17.1 (#286), DeepSeek-V2.5 (Sep 2024): 24.5 (#145)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Hard Prompts | 1251 | 1289 |
| CritPt | 0% | — |
| LiveBench Reasoning | 28.1% | — |
| DTBench | 56.7% | — |
| LiveBench Data Analysis | 48.5% | — |
| Epoch Capabilities Index | 127.15 | — |
| LiveBench | 43.5% | — |
Math DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 13.7 (#294), DeepSeek-V2.5 (Sep 2024): 33.3 (#177)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | 1244 | 1289 |
| OTIS Mock AIME 2024-2025 | 4.3% | — |
| Omni-MATH | 22.4% | — |
| LiveBench Math | 35.5% | — |
| MATH Level 5 | 46.4% | — |
| FrontierMath (Feb 2025 set) | 0.3% | — |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 18.7 (#276), DeepSeek-V2.5 (Sep 2024): 34.2 (#190)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | 1208 | 1266 |
| GPQA Diamond | 38.1% | — |
| MMLU-Pro | 60.5% | — |
| Confabulations | 36.7% | — |
| GPQA (HELM) | 36.3% | — |
| MMLU | 74.3% | — |
Multimodal Not comparable
Claude 3.5 Haiku: 26.3 (#113), DeepSeek-V2.5 (Sep 2024): —
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Vision | 1092 | — |
| GeoBench | 34% | — |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 40.2 (#213), DeepSeek-V2.5 (Sep 2024): 42.7 (#188)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | 1238 | 1273 |
| LMArena Chinese | 1229 | 1317 |
| LMArena French | 1262 | 1288 |
| LMArena German | 1237 | 1258 |
| LMArena Japanese | 1175 | 1228 |
| LMArena Korean | 1172 | 1209 |
| LMArena Russian | 1253 | 1289 |
| LMArena Spanish | 1261 | 1248 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 63.1 (#229), DeepSeek-V2.5 (Sep 2024): 67.7 (#189)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | 1241 | 1280 |
| LiveBench Instruction Following | 61.9% | — |
| IFEval | 79.2% | — |
Long Context DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 38.3 (#195), DeepSeek-V2.5 (Sep 2024): 39.5 (#169)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | 1261 | 1301 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
Claude 3.5 Haiku: 42.9 (#228), DeepSeek-V2.5 (Sep 2024): 50.0 (#182)
| Benchmark | Claude 3.5 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | 1255 | 1294 |
| LMArena Creative Writing | 1233 | 1285 |
| LMArena Multi-Turn | 1265 | 1297 |
| Short-Story Creative Writing | 73.5% | — |
| EQ-Bench Creative Writing | 1146 | — |
| WildBench | 76% | — |
| LiveBench Language | 35.4% | — |
Frequently asked questions
Is Claude 3.5 Haiku better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 28.9 on the Noometry Index.
Is Claude 3.5 Haiku or DeepSeek-V2.5 (Sep 2024) better for coding?
Claude 3.5 Haiku scores higher on coding benchmarks: 32.6 versus 31.3 in the Noometry coding category.
How many benchmarks do Claude 3.5 Haiku and DeepSeek-V2.5 (Sep 2024) share?
20 benchmarks have published results for both models. Claude 3.5 Haiku has 49 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.