Model comparison
DeepSeek-V3 vs Inkling
Inkling is the stronger model overall, scoring 44.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 6.4× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Last verified . 27 shared benchmarks.
DeepSeekSummary
- They share 27 benchmarks with published results for both. DeepSeek-V3 scores higher in 2 categories and Inkling in 6 categories; 7 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Inkling leads 40.4 to 20.5.
- The biggest single-benchmark swing is OTIS Mock AIME 2024-2025: 37.8% for DeepSeek-V3 and 88.9% for Inkling.
- DeepSeek-V3 is cheaper at $0.24 / $0.90 per million input/output tokens, against $1.87 / $4.68 for Inkling.
- DeepSeek-V3 accepts more context: 164K tokens versus 66K.
Side by side
| DeepSeek-V3 | Inkling | |
|---|---|---|
| Provider | DeepSeek | Thinking Machines Lab |
| Noometry Index | 39.5 | 44.1 |
| Released | 2024-12-26 | 2026-07-15 |
| Weights | Open | Open |
| Context window | 164K | 66K |
| Max output | 164K | 66K |
| Input $ / M tokens | $0.24 | $1.87 |
| Output $ / M tokens | $0.90 | $4.68 |
| Results tracked | 60 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3 leads
DeepSeek-V3: 42.3 (#106), Inkling: 34.5 (#234)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| SciCode | 35.8% | 47% |
| WeirdML | 36.1% | 32.3% |
| LMArena Coding | 1368 | 1464 |
| FrontierCode | — | 14% |
| Aider Polyglot | 55.1% | — |
| LMArena WebDev | — | 1413 |
| FrontierSWE | — | 4.1% |
| BigCodeBench Instruct | 50% | — |
| LiveBench Coding | 70.9% | — |
| BigCodeBench Complete | 62.2% | — |
| ALE-Bench | — | 946 |
| HumanEval+ | 86.6% | — |
| MBPP+ | 73% | — |
Agentic & Tool Use Not comparable
DeepSeek-V3: —, Inkling: 29.6 (#85)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| APEX-Agents | — | 33.8% |
| τ²-bench Banking | — | 25% |
| METR Time Horizons | 49.6% | — |
Reasoning Inkling leads
DeepSeek-V3: 20.5 (#236), Inkling: 40.4 (#56)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| SimpleBench | 27.2% | 50% |
| CritPt | 0% | 5.4% |
| LMArena Hard Prompts | 1365 | 1451 |
| DTBench | 64.8% | 87.5% |
| LMCA | 15.5% | 37.6% |
| Epoch Capabilities Index | 135.94 | 148.54 |
| ARC-AGI-2 | — | 36.5% |
| Kagi LLM Benchmark | 52.3% | — |
| ARC-AGI-1 | — | 79.5% |
| Chess Puzzles | — | 21% |
| LiveBench Reasoning | 65.8% | — |
| LiveBench Data Analysis | 60.9% | — |
| BIG-Bench Hard | 87.5% | — |
| ForecastBench | 59.1 | — |
| HellaSwag | 88.9% | — |
| LiveBench | 66.9% | — |
| PIQA | 84.7% | — |
| WinoGrande | 85.2% | — |
Math Too close to call
DeepSeek-V3: 32.1 (#219), Inkling: 31.3 (#225)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 37.8% | 88.9% |
| LMArena Math | 1373 | 1479 |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| FrontierMath Tier 4 | — | 4.9% |
| ProofBench | — | 0% |
| Omni-MATH | 40.3% | — |
| LiveBench Math | 73.5% | — |
| MATH Level 5 | 75.5% | — |
| FrontierMath (Feb 2025 set) | 1.7% | — |
Knowledge Inkling leads
DeepSeek-V3: 37.5 (#155), Inkling: 55.1 (#49)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| GPQA Diamond | 67.6% | 88.3% |
| LMArena Expert | 1351 | 1465 |
| SimpleQA Verified | — | 40.3% |
| MMLU-Pro | 72.3% | — |
| Confabulations | 26.1% | — |
| Vectara Hallucination Rate | 6.1% | — |
| GPQA (HELM) | 53.8% | — |
| ARC (AI2) Challenge | 95.3% | — |
| MMLU | 87.2% | — |
| TriviaQA | 82.9% | — |
Multilingual Inkling leads
DeepSeek-V3: 48.5 (#143), Inkling: 54.0 (#52)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| LMArena Non-English | 1358 | 1434 |
| LMArena Chinese | 1391 | 1490 |
| LMArena French | 1385 | 1458 |
| LMArena German | 1374 | 1446 |
| LMArena Japanese | 1333 | 1429 |
| LMArena Korean | 1319 | 1404 |
| LMArena Russian | 1373 | 1429 |
| LMArena Spanish | 1358 | 1448 |
Instruction Following Inkling leads
DeepSeek-V3: 72.8 (#130), Inkling: 75.1 (#71)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| LMArena Instruction Following | 1345 | 1426 |
| LiveBench Instruction Following | 81.5% | — |
| IFEval | 83.2% | — |
Long Context Inkling leads
DeepSeek-V3: 34.0 (#253), Inkling: 43.8 (#86)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| LMArena Longer Query | 1352 | 1434 |
| Fiction.LiveBench | 50% | — |
Writing & Preference Inkling leads
DeepSeek-V3: 57.4 (#130), Inkling: 65.2 (#51)
| Benchmark | DeepSeek-V3 | Inkling |
|---|---|---|
| LMArena Text | 1375 | 1441 |
| LMArena Creative Writing | 1364 | 1387 |
| EQ-Bench Creative Writing | 1472 | 1611 |
| LMArena Multi-Turn | 1389 | 1436 |
| Short-Story Creative Writing | 77% | — |
| WildBench | 83% | — |
| EQ-Bench 4 | — | 1226 |
| LiveBench Language | 49.1% | — |
Frequently asked questions
Is DeepSeek-V3 better than Inkling?
Inkling is the stronger model overall, scoring 44.1 to 39.5 on the Noometry Index. DeepSeek-V3 costs 6.4× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3 or Inkling?
DeepSeek-V3 is cheaper. It lists at $0.24 per million input tokens and $0.90 per million output tokens; Inkling lists at $1.87 and $4.68.
Is DeepSeek-V3 or Inkling better for coding?
DeepSeek-V3 scores higher on coding benchmarks: 42.3 versus 34.5 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3 does, with 164K tokens against 66K.
How many benchmarks do DeepSeek-V3 and Inkling share?
27 benchmarks have published results for both models. DeepSeek-V3 has 60 scored results on Noometry and Inkling has 41.