Model comparison
DeepSeek-V3 vs DeepSeek-V3.1-Terminus
DeepSeek-V3.1-Terminus is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index.
Last verified . 15 shared benchmarks.
DeepSeekSummary
- They share 15 benchmarks with published results for both. DeepSeek-V3 scores higher in 1 category and DeepSeek-V3.1-Terminus in 6 categories; 6 gaps are clear of the uncertainty.
- The widest gap is in long context, where DeepSeek-V3.1-Terminus leads 43.4 to 34.0.
- The biggest single-benchmark swing is DTBench: 64.8% for DeepSeek-V3 and 81.3% for DeepSeek-V3.1-Terminus.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $0.27 / $1 for DeepSeek-V3.1-Terminus.
Side by side
| DeepSeek-V3 | DeepSeek-V3.1-Terminus | |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Noometry Index | 39.5 | 43.1 |
| Released | 2024-12-26 | 2025-09-22 |
| Weights | Open | Open |
| Context window | 164K | 164K |
| Max output | 164K | 147K |
| Input $ / M tokens | $0.24 | $0.27 |
| Output $ / M tokens | $0.90 | $1 |
| Results tracked | 60 | 16 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding Too close to call
DeepSeek-V3: 42.3 (#106), DeepSeek-V3.1-Terminus: 42.0 (#113)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| SciCode | 35.8% | 40.6% |
| LMArena Coding | 1368 | 1426 |
| Aider Polyglot | 55.1% | — |
| WeirdML | 36.1% | — |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 745.17 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, DeepSeek-V3.1-Terminus: —
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| METR Time Horizons | 49.6% | — |
Reasoning DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 20.5 (#236), DeepSeek-V3.1-Terminus: 26.4 (#133)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| Kagi LLM Benchmark | 52.3% | 57.4% |
| CritPt | 0% | 1.7% |
| LMArena Hard Prompts | 1365 | 1426 |
| DTBench | 64.8% | 81.3% |
| LMCA | 15.5% | 28.6% |
| SimpleBench | 27.2% | — |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| 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 DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 32.1 (#219), DeepSeek-V3.1-Terminus: 38.5 (#137)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| LMArena Math | 1373 | 1402 |
| OTIS Mock AIME 2024-2025 | 37.8% | — |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge Not comparable
DeepSeek-V3: 37.5 (#155), DeepSeek-V3.1-Terminus: —
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| GPQA Diamond | 67.6% | — |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| LMArena Expert | 1351 | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 48.5 (#143), DeepSeek-V3.1-Terminus: 52.1 (#92)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| LMArena Non-English | 1358 | 1407 |
| LMArena Russian | 1373 | 1436 |
| LMArena Chinese | 1391 | — |
| LMArena French | 1385 | — |
| LMArena German | 1374 | — |
| LMArena Japanese | 1333 | — |
| LMArena Korean | 1319 | — |
| LMArena Spanish | 1358 | — |
Instruction Following DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 72.8 (#130), DeepSeek-V3.1-Terminus: 74.0 (#106)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| LMArena Instruction Following | 1345 | 1404 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 34.0 (#253), DeepSeek-V3.1-Terminus: 43.4 (#97)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| LMArena Longer Query | 1352 | 1421 |
| Fiction.LiveBench | 50% | — |
Writing & Preference DeepSeek-V3.1-Terminus leads
DeepSeek-V3: 57.4 (#130), DeepSeek-V3.1-Terminus: 61.0 (#92)
| Benchmark | DeepSeek-V3 | DeepSeek-V3.1-Terminus |
|---|---|---|
| LMArena Text | 1375 | 1419 |
| LMArena Creative Writing | 1364 | 1403 |
| LMArena Multi-Turn | 1389 | 1411 |
| Short-Story Creative Writing | 77% | — |
| EQ-Bench Creative Writing | 1472 | — |
| WildBench | 83% | — |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than DeepSeek-V3.1-Terminus?
DeepSeek-V3.1-Terminus is the stronger model overall, scoring 43.1 to 39.5 on the Noometry Index.
Which is cheaper, DeepSeek-V3 or DeepSeek-V3.1-Terminus?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; DeepSeek-V3.1-Terminus lists at $0.27 and $1.
Is DeepSeek-V3 or DeepSeek-V3.1-Terminus better for coding?
They score almost the same on coding (42.3 vs 42.0); test both on your own repository before choosing.
Which has the bigger context window?
Both accept 164K tokens.
How many benchmarks do DeepSeek-V3 and DeepSeek-V3.1-Terminus share?
15 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and DeepSeek-V3.1-Terminus has 16.