Model comparison
DeepSeek-V3.2-Exp vs Inkling
DeepSeek-V3.2-Exp and Inkling score almost the same on the Noometry Index (44.3 vs 44.1), so choose on price, context window or the category you care about most.
Last verified . 32 shared benchmarks.
DeepSeekSummary
- They share 32 benchmarks with published results for both. DeepSeek-V3.2-Exp scores higher in 4 categories and Inkling in 5 categories; 8 gaps are clear of the uncertainty.
- The widest gap is in reasoning, where Inkling leads 40.4 to 22.1.
- The biggest single-benchmark swing is ARC-AGI-2: 4% for DeepSeek-V3.2-Exp and 36.5% for Inkling.
- DeepSeek-V3.2-Exp is cheaper at $0.26 / $0.38 per million input/output tokens, against $1.87 / $4.68 for Inkling.
- DeepSeek-V3.2-Exp accepts more context: 164K tokens versus 66K.
Side by side
| DeepSeek-V3.2-Exp | Inkling | |
|---|---|---|
| Provider | DeepSeek | Thinking Machines Lab |
| Noometry Index | 44.3 | 44.1 |
| Released | 2025-09-29 | 2026-07-15 |
| Weights | Open | Open |
| Context window | 164K | 66K |
| Max output | 66K | 66K |
| Input $ / M tokens | $0.26 | $1.87 |
| Output $ / M tokens | $0.38 | $4.68 |
| Results tracked | 49 | 41 |
Sponsored placements are available on pages like this one. Advertise on Noometry
Category by category
Coding DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 46.5 (#65), Inkling: 34.5 (#234)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| LMArena WebDev | 1362 | 1413 |
| SciCode | 38.9% | 47% |
| WeirdML | 39.5% | 32.3% |
| LMArena Coding | 1454 | 1464 |
| FrontierCode | — | 14% |
| SWE-bench Verified (bash only) | 70% | — |
| Aider Polyglot | 74.2% | — |
| SWE-bench Multilingual | 59% | — |
| FrontierSWE | — | 4.1% |
| ALE-Bench | — | 946 |
Agentic & Tool Use DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 32.7 (#59), Inkling: 29.6 (#85)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| APEX-Agents | 21.3% | 33.8% |
| Terminal-Bench | 39.6% | — |
| Berkeley Function Calling Leaderboard | 56.7% | — |
| TheAgentCompany | 42.9% | — |
| τ²-bench Banking | — | 25% |
| Vending-Bench 2 | 1,034 | — |
Reasoning Inkling leads
DeepSeek-V3.2-Exp: 22.1 (#208), Inkling: 40.4 (#56)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| ARC-AGI-2 | 4% | 36.5% |
| ARC-AGI-1 | 57% | 79.5% |
| CritPt | 2.9% | 5.4% |
| Chess Puzzles | 14% | 21% |
| LMArena Hard Prompts | 1434 | 1451 |
| DTBench | 87.7% | 87.5% |
| LMCA | 29.1% | 37.6% |
| Epoch Capabilities Index | 146.27 | 148.54 |
| SimpleBench | — | 50% |
| Kagi LLM Benchmark | 52.2% | — |
| NYT Connections (extended) | 36.7% | — |
| Thematic Generalization | 65% | — |
Math DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 41.7 (#87), Inkling: 31.3 (#225)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| OTIS Mock AIME 2024-2025 | 87.8% | 88.9% |
| ProofBench | 8% | 0% |
| LMArena Math | 1435 | 1479 |
| FrontierMath (Tiers 1-3) | — | 33.3% |
| FrontierMath Tier 4 | — | 4.9% |
| MathArena Final-Answer Competitions | 57.7% | — |
| FrontierMath (Feb 2025 set) | 22.1% | — |
| FrontierMath Tier 4 (v1) | 2.1% | — |
Knowledge Inkling leads
DeepSeek-V3.2-Exp: 51.7 (#66), Inkling: 55.1 (#49)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| GPQA Diamond | 83.4% | 88.3% |
| LMArena Expert | 1436 | 1465 |
| SimpleQA Verified | — | 40.3% |
| Vectara Hallucination Rate | 5.3% | — |
Multilingual Inkling leads
DeepSeek-V3.2-Exp: 52.2 (#90), Inkling: 54.0 (#52)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| LMArena Non-English | 1409 | 1434 |
| LMArena Chinese | 1461 | 1490 |
| LMArena French | 1433 | 1458 |
| LMArena German | 1440 | 1446 |
| LMArena Japanese | 1374 | 1429 |
| LMArena Korean | 1371 | 1404 |
| LMArena Russian | 1424 | 1429 |
| LMArena Spanish | 1440 | 1448 |
Instruction Following Too close to call
DeepSeek-V3.2-Exp: 74.5 (#93), Inkling: 75.1 (#71)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| LMArena Instruction Following | 1413 | 1426 |
Long Context DeepSeek-V3.2-Exp leads
DeepSeek-V3.2-Exp: 47.6 (#16), Inkling: 43.8 (#86)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| LMArena Longer Query | 1428 | 1434 |
| Fiction.LiveBench | 83.3% | — |
| CL-bench | 13.2% | — |
| CL-bench Life | 9.5% | — |
Writing & Preference Inkling leads
DeepSeek-V3.2-Exp: 62.4 (#77), Inkling: 65.2 (#51)
| Benchmark | DeepSeek-V3.2-Exp | Inkling |
|---|---|---|
| LMArena Text | 1425 | 1441 |
| LMArena Creative Writing | 1403 | 1387 |
| EQ-Bench Creative Writing | 1515 | 1611 |
| LMArena Multi-Turn | 1427 | 1436 |
| EQ-Bench 4 | — | 1226 |
Frequently asked questions
Is DeepSeek-V3.2-Exp better than Inkling?
DeepSeek-V3.2-Exp and Inkling score almost the same on the Noometry Index (44.3 vs 44.1), so choose on price, context window or the category you care about most.
Which is cheaper, DeepSeek-V3.2-Exp or Inkling?
DeepSeek-V3.2-Exp is cheaper. It lists at $0.26 per million input tokens and $0.38 per million output tokens; Inkling lists at $1.87 and $4.68.
Is DeepSeek-V3.2-Exp or Inkling better for coding?
DeepSeek-V3.2-Exp scores higher on coding benchmarks: 46.5 versus 34.5 in the Noometry coding category.
Which has the bigger context window?
DeepSeek-V3.2-Exp does, with 164K tokens against 66K.
How many benchmarks do DeepSeek-V3.2-Exp and Inkling share?
32 benchmarks have published results for both models. DeepSeek-V3.2-Exp has 49 scored results on Noometry and Inkling has 41.