Model comparison
DeepSeek-V3.2-Speciale vs Inkling
Inkling is the stronger model overall, scoring 44.1 to 39.7 on the Noometry Index. DeepSeek-V3.2-Speciale costs 3.0× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Last verified . 3 shared benchmarks.
DeepSeekSummary
- They share 3 benchmarks with published results for both. DeepSeek-V3.2-Speciale scores higher in 1 category and Inkling in 2 categories; 3 gaps are clear of the uncertainty.
- The widest gap is in writing & preference, where Inkling leads 65.2 to 46.0.
- The biggest single-benchmark swing is WeirdML: 46.7% for DeepSeek-V3.2-Speciale and 32.3% for Inkling.
- DeepSeek-V3.2-Speciale is cheaper at $0.58 / $1.68 per million input/output tokens, against $1.87 / $4.68 for Inkling.
- DeepSeek-V3.2-Speciale accepts more context: 128K tokens versus 66K.
Side by side
| DeepSeek-V3.2-Speciale | Inkling | |
|---|---|---|
| Provider | DeepSeek | Thinking Machines Lab |
| Noometry Index | 39.7 | 44.1 |
| Released | 2025-12-01 | 2026-07-15 |
| Weights | Open | Open |
| Context window | 128K | 66K |
| Max output | 128K | 66K |
| Input $ / M tokens | $0.58 | $1.87 |
| Output $ / M tokens | $1.68 | $4.68 |
| Results tracked | 3 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Speciale leads
DeepSeek-V3.2-Speciale: 40.4 (#140), Inkling: 34.5 (#234)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| WeirdML | 46.7% | 32.3% |
| FrontierCode | — | 14% |
| LMArena WebDev | — | 1413 |
| FrontierSWE | — | 4.1% |
| SciCode | — | 47% |
| LMArena Coding | — | 1464 |
| ALE-Bench | — | 946 |
Agentic & Tool Use Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 29.6 (#85)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| APEX-Agents | — | 33.8% |
| τ²-bench Banking | — | 25% |
Reasoning Inkling leads
DeepSeek-V3.2-Speciale: 32.9 (#73), Inkling: 40.4 (#56)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| SimpleBench | 52.6% | 50% |
| ARC-AGI-2 | — | 36.5% |
| ARC-AGI-1 | — | 79.5% |
| CritPt | — | 5.4% |
| Chess Puzzles | — | 21% |
| LMArena Hard Prompts | — | 1451 |
| DTBench | — | 87.5% |
| LMCA | — | 37.6% |
| Epoch Capabilities Index | — | 148.54 |
Math Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 31.3 (#225)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| FrontierMath (Tiers 1-3) | — | 33.3% |
| FrontierMath Tier 4 | — | 4.9% |
| OTIS Mock AIME 2024-2025 | — | 88.9% |
| ProofBench | — | 0% |
| LMArena Math | — | 1479 |
Knowledge Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 55.1 (#49)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| GPQA Diamond | — | 88.3% |
| SimpleQA Verified | — | 40.3% |
| LMArena Expert | — | 1465 |
Multilingual Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 54.0 (#52)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| LMArena Non-English | — | 1434 |
| LMArena Chinese | — | 1490 |
| LMArena French | — | 1458 |
| LMArena German | — | 1446 |
| LMArena Japanese | — | 1429 |
| LMArena Korean | — | 1404 |
| LMArena Russian | — | 1429 |
| LMArena Spanish | — | 1448 |
Instruction Following Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 75.1 (#71)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| LMArena Instruction Following | — | 1426 |
Long Context Not comparable
DeepSeek-V3.2-Speciale: —, Inkling: 43.8 (#86)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| LMArena Longer Query | — | 1434 |
Writing & Preference Inkling leads
DeepSeek-V3.2-Speciale: 46.0 (#222), Inkling: 65.2 (#51)
| Benchmark | DeepSeek-V3.2-Speciale | Inkling |
|---|---|---|
| EQ-Bench Creative Writing | 1276 | 1611 |
| LMArena Text | — | 1441 |
| LMArena Creative Writing | — | 1387 |
| EQ-Bench 4 | — | 1226 |
| LMArena Multi-Turn | — | 1436 |
Frequently asked questions
Is DeepSeek-V3.2-Speciale better than Inkling?
Inkling is the stronger model overall, scoring 44.1 to 39.7 on the Noometry Index. DeepSeek-V3.2-Speciale costs 3.0× less per token, which makes it the better buy when Inkling's lead doesn't matter for your workload.
Which is cheaper, DeepSeek-V3.2-Speciale or Inkling?
DeepSeek-V3.2-Speciale is cheaper. It lists at $0.58 per million input tokens and $1.68 per million output tokens; Inkling lists at $1.87 and $4.68.
Is DeepSeek-V3.2-Speciale or Inkling better for coding?
DeepSeek-V3.2-Speciale scores higher on coding benchmarks: 40.4 versus 34.5 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.2-Speciale does, with 128K tokens against 66K.
How many benchmarks do DeepSeek-V3.2-Speciale and Inkling share?
3 benchmarks have published results for both models. DeepSeek-V3.2-Speciale has 3 scored results on Noometry and Inkling has 41.