Model comparison
Claude 3 Haiku vs DeepSeek-V2.5 (Sep 2024)
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 25.7 on the Noometry Index.
Last verified . 21 shared benchmarks.
DeepSeekSummary
- They share 21 benchmarks with published results for both. Claude 3 Haiku scores higher in 0 categories and DeepSeek-V2.5 (Sep 2024) in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek-V2.5 (Sep 2024) leads 33.3 to 9.2.
- The biggest single-benchmark swing is BigCodeBench Instruct: 39.4% for Claude 3 Haiku and 48.6% 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 Haiku | DeepSeek-V2.5 (Sep 2024) | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 25.7 | 37.0 |
| Released | 2024-03-07 | 2024-09-06 |
| Weights | Proprietary | Open |
| Context window | — | — |
| Max output | — | — |
| Input $ / M tokens | — | — |
| Output $ / M tokens | — | — |
| Results tracked | 37 | 22 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 26.1 (#320), DeepSeek-V2.5 (Sep 2024): 31.3 (#274)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| BigCodeBench Instruct | 39.4% | 48.6% |
| LMArena Coding | 1199 | 1310 |
| BigCodeBench Complete | 50.1% | 53.2% |
| HumanEval+ | 68.9% | 83.5% |
| MBPP+ | 68.8% | 74.1% |
| Aider Polyglot | — | 17.8% |
| WeirdML | 9.8% | — |
| CadEval | 12% | — |
Reasoning DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 15.7 (#303), DeepSeek-V2.5 (Sep 2024): 24.5 (#145)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Hard Prompts | 1174 | 1289 |
| Kagi LLM Benchmark | 34.2% | — |
| DTBench | 50.1% | — |
| LMCA | 8.8% | — |
| Epoch Capabilities Index | 118.3 | — |
| ForecastBench | 53.2 | — |
| WinoGrande | 74.2% | — |
Math DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 9.2 (#313), DeepSeek-V2.5 (Sep 2024): 33.3 (#177)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Math | 1188 | 1289 |
| OTIS Mock AIME 2024-2025 | 1.8% | — |
| MATH Level 5 | 14.9% | — |
Knowledge DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 17.3 (#280), DeepSeek-V2.5 (Sep 2024): 34.2 (#190)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Expert | 1148 | 1266 |
| GPQA Diamond | 36.3% | — |
| Confabulations | 34.2% | — |
| MMLU | 73.8% | — |
Multimodal Not comparable
Claude 3 Haiku: 23.1 (#123), DeepSeek-V2.5 (Sep 2024): —
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Vision | 950 | — |
| ScienceQA | 72% | — |
Multilingual DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 36.1 (#238), DeepSeek-V2.5 (Sep 2024): 42.7 (#188)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Non-English | 1178 | 1273 |
| LMArena Chinese | 1155 | 1317 |
| LMArena French | 1194 | 1288 |
| LMArena German | 1174 | 1258 |
| LMArena Japanese | 1102 | 1228 |
| LMArena Korean | 1108 | 1209 |
| LMArena Russian | 1204 | 1289 |
| LMArena Spanish | 1165 | 1248 |
Instruction Following DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 61.5 (#242), DeepSeek-V2.5 (Sep 2024): 67.7 (#189)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Instruction Following | 1174 | 1280 |
Long Context DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 36.1 (#232), DeepSeek-V2.5 (Sep 2024): 39.5 (#169)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Longer Query | 1191 | 1301 |
Writing & Preference DeepSeek-V2.5 (Sep 2024) leads
Claude 3 Haiku: 29.8 (#286), DeepSeek-V2.5 (Sep 2024): 50.0 (#182)
| Benchmark | Claude 3 Haiku | DeepSeek-V2.5 (Sep 2024) |
|---|---|---|
| LMArena Text | 1195 | 1294 |
| LMArena Creative Writing | 1157 | 1285 |
| LMArena Multi-Turn | 1190 | 1297 |
| EQ-Bench Creative Writing | 717 | — |
Frequently asked questions
Is Claude 3 Haiku better than DeepSeek-V2.5 (Sep 2024)?
DeepSeek-V2.5 (Sep 2024) is the stronger model overall, scoring 37.0 to 25.7 on the Noometry Index.
Is Claude 3 Haiku or DeepSeek-V2.5 (Sep 2024) better for coding?
DeepSeek-V2.5 (Sep 2024) scores higher on coding benchmarks: 31.3 versus 26.1 in the Noometry coding category.
How many benchmarks do Claude 3 Haiku and DeepSeek-V2.5 (Sep 2024) share?
21 benchmarks have published results for both models. Claude 3 Haiku has 37 scored results on Noometry and DeepSeek-V2.5 (Sep 2024) has 22.