Model comparison
DeepSeek-V3 vs Nova Premier 1.0
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 38.3 on the Noometry Index.
Last verified . 6 shared benchmarks.
DeepSeekSummary
- They share 6 benchmarks with published results for both. DeepSeek-V3 scores higher in 3 categories and Nova Premier 1.0 in 2 categories; 5 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where DeepSeek-V3 leads 57.4 to 51.0.
- The biggest single-benchmark swing is Kagi LLM Benchmark: 52.3% for DeepSeek-V3 and 44.8% for Nova Premier 1.0.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $2.50 / $12.50 for Nova Premier 1.0.
- Nova Premier 1.0 accepts more context: 1M tokens versus 164K.
- DeepSeek-V3 has downloadable open weights; the other is API-only.
Side by side
| DeepSeek-V3 | Nova Premier 1.0 | |
|---|---|---|
| Provider | DeepSeek | Amazon |
| Noometry Index | 39.5 | 38.3 |
| Released | 2024-12-26 | 2025-04-30 |
| Weights | Open | Proprietary |
| Context window | 164K | 1M |
| Max output | 164K | 10K |
| Input $ / M tokens | $0.24 | $2.50 |
| Output $ / M tokens | $0.90 | $12.50 |
| Results tracked | 60 | 6 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Not comparable
DeepSeek-V3: 42.3 (#106), Nova Premier 1.0: —
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| Aider Polyglot | 55.1% | — |
| SciCode | 35.8% | — |
| WeirdML | 36.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| LMArena Coding | 1368 | — |
| BigCodeBench Complete | 62.2% | — |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Nova Premier 1.0: —
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| METR Time Horizons | 49.6% | — |
Reasoning Nova Premier 1.0 leads
DeepSeek-V3: 20.5 (#236), Nova Premier 1.0: 23.0 (#185)
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 44.8% |
| SimpleBench | 27.2% | — |
| CritPt | 0% | — |
| LiveBench Reasoning | 65.8% | — |
| LMArena Hard Prompts | 1365 | — |
| DTBench | 64.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| LMCA | 15.5% | — |
| BIG-Bench Hard | 87.5% | — |
| Epoch Capabilities Index | 135.94 | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math Nova Premier 1.0 leads
DeepSeek-V3: 32.1 (#219), Nova Premier 1.0: 33.8 (#199)
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| Omni-MATH | 40.3% | 35% |
| OTIS Mock AIME 2024-2025 | 37.8% | — |
| LiveBench Math | 73.5% | — |
| LMArena Math | 1373 | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge DeepSeek-V3 leads
DeepSeek-V3: 37.5 (#155), Nova Premier 1.0: 35.8 (#180)
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| MMLU-Pro | 72.3% | 72.6% |
| GPQA (HELM) | 53.8% | 51.8% |
| GPQA Diamond | 67.6% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| LMArena Expert | 1351 | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual Not comparable
DeepSeek-V3: 48.5 (#143), Nova Premier 1.0: —
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| LMArena Non-English | 1358 | — |
| LMArena Chinese | 1391 | — |
| LMArena French | 1385 | — |
| LMArena German | 1374 | — |
| LMArena Japanese | 1333 | — |
| LMArena Korean | 1319 | — |
| LMArena Russian | 1373 | — |
| LMArena Spanish | 1358 | — |
Instruction Following DeepSeek-V3 leads
DeepSeek-V3: 72.8 (#130), Nova Premier 1.0: 66.8 (#204)
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| IFEval | 83.2% | 80.3% |
| LiveBench Instruction Following | 81.5% | — |
| LMArena Instruction Following | 1345 | — |
Long Context Not comparable
DeepSeek-V3: 34.0 (#253), Nova Premier 1.0: —
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| Fiction.LiveBench | 50% | — |
| LMArena Longer Query | 1352 | — |
Writing & Preference DeepSeek-V3 leads
DeepSeek-V3: 57.4 (#130), Nova Premier 1.0: 51.0 (#178)
| Benchmark | DeepSeek-V3 | Nova Premier 1.0 |
|---|---|---|
| WildBench | 83% | 78.8% |
| LMArena Text | 1375 | — |
| LMArena Creative Writing | 1364 | — |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| LMArena Multi-Turn | 1389 | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Nova Premier 1.0?
DeepSeek-V3 is the stronger model overall, scoring 39.5 to 38.3 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or Nova Premier 1.0?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Nova Premier 1.0 lists at $2.50 and $12.50.
Which has the bigger context window?
Nova Premier 1.0 does, with 1M tokens against 164K.
How many benchmarks do DeepSeek-V3 and Nova Premier 1.0 share?
6 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Nova Premier 1.0 has 6.