Model comparison
Claude 3 Haiku vs DeepSeek V4 Flash
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 25.7 on the Noometry Index.
Last verified . 25 shared benchmarks.
DeepSeekSummary
- They share 25 benchmarks with published results for both. Claude 3 Haiku scores higher in 0 categories and DeepSeek V4 Flash in 8 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in math, where DeepSeek V4 Flash leads 60.0 to 9.2.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 1.8% for Claude 3 Haiku and 94.4% for DeepSeek V4 Flash.
- DeepSeek V4 Flash has downloadable open weights; the other is API-only.
Side by side
| Claude 3 Haiku | DeepSeek V4 Flash | |
|---|---|---|
| Provider | Anthropic | DeepSeek |
| Noometry Index | 25.7 | 53.6 |
| Released | 2024-03-07 | 2026-04-24 |
| Weights | Proprietary | Open |
| Context window | — | 1M |
| Max output | — | 393K |
| Input $ / M tokens | — | $0.15 |
| Output $ / M tokens | — | $0.60 |
| Results tracked | 37 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek V4 Flash leads
Claude 3 Haiku: 26.1 (#320), DeepSeek V4 Flash: 47.6 (#56)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| WeirdML | 9.8% | 63% |
| LMArena Coding | 1199 | 1457 |
| FrontierCode | — | 18.8% |
| LMArena WebDev | — | 1581 |
| SciCode | — | 49.9% |
| BigCodeBench Instruct | 39.4% | — |
| BigCodeBench Complete | 50.1% | — |
| CadEval | 12% | — |
| ALE-Bench | — | 1,306 |
| HumanEval+ | 68.9% | — |
| MBPP+ | 68.8% | — |
Reasoning DeepSeek V4 Flash leads
Claude 3 Haiku: 15.7 (#303), DeepSeek V4 Flash: 53.7 (#25)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| Kagi LLM Benchmark | 34.2% | 52.2% |
| LMArena Hard Prompts | 1174 | 1443 |
| DTBench | 50.1% | 90.9% |
| LMCA | 8.8% | 41.7% |
| Epoch Capabilities Index | 118.3 | 154.51 |
| ARC-AGI-2 | — | 61.4% |
| SimpleBench | — | 61.1% |
| NYT Connections (extended) | — | 89.6% |
| ARC-AGI-1 | — | 89% |
| CritPt | — | 16.6% |
| Chess Puzzles | — | 33% |
| Mystery Game Puzzles | — | 34% |
| ForecastBench | 53.2 | — |
| WinoGrande | 74.2% | — |
Math DeepSeek V4 Flash leads
Claude 3 Haiku: 9.2 (#313), DeepSeek V4 Flash: 60.0 (#31)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 1.8% | 94.4% |
| LMArena Math | 1188 | 1427 |
| FrontierMath (Tiers 1-3) | — | 57.5% |
| FrontierMath Tier 4 | — | 24.4% |
| MathArena Final-Answer Competitions | — | 76.5% |
| ProofBench | — | 56% |
| MATH Level 5 | 14.9% | — |
Knowledge DeepSeek V4 Flash leads
Claude 3 Haiku: 17.3 (#280), DeepSeek V4 Flash: 55.5 (#43)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| GPQA Diamond | 36.3% | 91% |
| LMArena Expert | 1148 | 1441 |
| SimpleQA Verified | — | 33.6% |
| Confabulations | 34.2% | — |
| MMLU | 73.8% | — |
Multimodal Not comparable
Claude 3 Haiku: 23.1 (#123), DeepSeek V4 Flash: —
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| LMArena Vision | 950 | — |
| ScienceQA | 72% | — |
Multilingual DeepSeek V4 Flash leads
Claude 3 Haiku: 36.1 (#238), DeepSeek V4 Flash: 53.2 (#71)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| LMArena Non-English | 1178 | 1420 |
| LMArena Chinese | 1155 | 1467 |
| LMArena French | 1194 | 1438 |
| LMArena German | 1174 | 1418 |
| LMArena Japanese | 1102 | 1403 |
| LMArena Korean | 1108 | 1383 |
| LMArena Russian | 1204 | 1428 |
| LMArena Spanish | 1165 | 1437 |
Instruction Following DeepSeek V4 Flash leads
Claude 3 Haiku: 61.5 (#242), DeepSeek V4 Flash: 75.1 (#76)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| LMArena Instruction Following | 1174 | 1421 |
Long Context DeepSeek V4 Flash leads
Claude 3 Haiku: 36.1 (#232), DeepSeek V4 Flash: 43.8 (#81)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| LMArena Longer Query | 1191 | 1435 |
Writing & Preference DeepSeek V4 Flash leads
Claude 3 Haiku: 29.8 (#286), DeepSeek V4 Flash: 64.0 (#59)
| Benchmark | Claude 3 Haiku | DeepSeek V4 Flash |
|---|---|---|
| LMArena Text | 1195 | 1432 |
| LMArena Creative Writing | 1157 | 1403 |
| EQ-Bench Creative Writing | 717 | 1559 |
| LMArena Multi-Turn | 1190 | 1449 |
Frequently asked questions
Is Claude 3 Haiku better than DeepSeek V4 Flash?
DeepSeek V4 Flash is the stronger model overall, scoring 53.6 to 25.7 on the Noometry Index.
Is Claude 3 Haiku or DeepSeek V4 Flash better for coding?
DeepSeek V4 Flash scores higher on coding benchmarks: 47.6 versus 26.1 in the Noometry coding category.
How many benchmarks do Claude 3 Haiku and DeepSeek V4 Flash share?
25 benchmarks have published results for both models. Claude 3 Haiku has 37 scored results on Noometry and DeepSeek V4 Flash has 41.