Inkling 2 Small vs DeepSeek-V4.1-Flash-1017
Both models appear in the same evaluation table on the official model card, which is the only place they are measured under identical conditions and reported by the same party.
Across 48 benchmarks where both models report a score, Inkling 2 Small is higher on 27 and DeepSeek-V4.1-Flash-1017 is higher on 21 . Inkling 2 Small has 594B total parameters with 17B active, and is priced at $1 per million output tokens on Tinker. Its largest margin is on DesignArena Game Dev. Its largest deficit is on GDPVal-AA v2.1.
Head to head
- Position
- Inkling 2 Small: the subject of this site · DeepSeek-V4.1-Flash-1017: open weights, smaller class
- Bechmarks scored
- 48 shared rows, out of 71 total
- Record
- 27 wins · 21 losses · 0 ties
Largest differences
| Benchmark | Inkling 2 Small | DeepSeek-V4.1-Flash-1017 | Gap |
|---|---|---|---|
| DesignArena Game Dev · Visual Creation | 1,428 | 1,261 | + 167 |
| GDPVal-AA v2.1 · Knowledge Work | 1,630 | 1,746 | -116 |
| AA-Briefcase v1.1 · Knowledge Work | 1,527 | 1,570 | -43 |
| DesignArena Website · Visual Creation | 1,340 | 1,297 | + 43 |
| DesignArena 3D Design · Visual Creation | 1,401 | 1,361 | + 40 |
| AA-Omniscience v1.1 · Factuality | 73.3% | 35.9% | + 37.4 |
| BabyVision · Visual Perception | 68.6% | 92.5% | -23.9 |
| AutomationBench v1.0.6 · Enterprise Operations | 45.6% | 66.6% | -21 |
| Terminal Bench v4.0 · Coding | 13.1% | 33.8% | -20.7 |
| Chartography · Visual Reasoning | 36.7% | 17.9% | + 18.8 |
Every shared benchmark
| Benchmark | Inkling 2 Small | DeepSeek-V4.1-Flash-1017 | Leader |
|---|---|---|---|
| Third Party Index | |||
| AA Index (v5.0) | 30.6 | 32.6 | DeepSeek-V4.1-Flash-1017 |
| Vals Index (v2.2) | 56.1% | 61.3% | DeepSeek-V4.1-Flash-1017 |
| Coding | |||
| FrontierCode v2.0 | 25.4% | 28.5% | DeepSeek-V4.1-Flash-1017 |
| DeepSWE v1.1 | 72.1% | 77% | DeepSeek-V4.1-Flash-1017 |
| Terminal Bench Science v0.1 | 5.7% | 12.9% | DeepSeek-V4.1-Flash-1017 |
| Terminal Bench v4.0 | 13.1% | 33.8% | DeepSeek-V4.1-Flash-1017 |
| Terminal Bench 2.1 · Inkling 2 Small: Pi harness | 85.1% | 78.4% | Inkling 2 Small |
| Vibe Code Bench v1.1 | 81.6% | 87.7% | DeepSeek-V4.1-Flash-1017 |
| SWE-Marathon v1.1 | 38.8% | 40% | DeepSeek-V4.1-Flash-1017 |
| SWEBench Multilingual | 81.2% | 82.9% | DeepSeek-V4.1-Flash-1017 |
| SWEBench Verified | 82.2% | 84.6% | DeepSeek-V4.1-Flash-1017 |
| Autoresearch | |||
| PostTrainBench v1.2 | 31.9% | 30.5% | Inkling 2 Small |
| PMPP-Hard · AUC (task-hours, 0–10h) | 494.27 | 497.36 | DeepSeek-V4.1-Flash-1017 |
| Search and Tools | |||
| MCP Atlas · Full / Public | 85.4% | 83.2% | Inkling 2 Small |
| Knowledge Work | |||
| AA-Briefcase v1.1 | 1,527 | 1,570 | DeepSeek-V4.1-Flash-1017 |
| GDPVal-AA v2.1 | 1,630 | 1,746 | DeepSeek-V4.1-Flash-1017 |
| Enterprise Operations | |||
| AutomationBench v1.0.6 | 45.6% | 66.6% | DeepSeek-V4.1-Flash-1017 |
| Finance Work | |||
| Finance Agent v2.1 | 67.5% | 62.9% | Inkling 2 Small |
| Excel Modeling Benchmark v1.1 | 69.5% | 69.2% | Inkling 2 Small |
| Tax Agent Bench | 65.8% | 74.9% | DeepSeek-V4.1-Flash-1017 |
| DAYJOB (Finance) | 14.3% | 12.5% | Inkling 2 Small |
| Apex-Accounting v1.1 · pass@1 / mean score | 14.5% | 12.5% | Inkling 2 Small |
| Diligence Bench · Best harness | 58.1% | 50.4% | Inkling 2 Small |
| Legal Work | |||
| Harvey LAB v1.1 | 21.7% | 14.2% | Inkling 2 Small |
| Legal Research Bench | 46.6% | 45.2% | Inkling 2 Small |
| Visual Creation | |||
| DesignArena Website | 1,340 | 1,297 | Inkling 2 Small |
| DesignArena Game Dev | 1,428 | 1,261 | Inkling 2 Small |
| DesignArena 3D Design | 1,401 | 1,361 | Inkling 2 Small |
| Code Arena | 1,653 | 1,665 | DeepSeek-V4.1-Flash-1017 |
| Visual Perception | |||
| PerceptionBench | 64.7% | 47.7% | Inkling 2 Small |
| Humanity’s Sixth Sense | 33.3% | 30.1% | Inkling 2 Small |
| BabyVision · original / with Python | 68.6% | 92.5% | DeepSeek-V4.1-Flash-1017 |
| Visual Reasoning | |||
| GDP.pdf | 25.2% | 17.2% | Inkling 2 Small |
| MMMU Pro (Standard 10) | 81.6% | 80.6% | Inkling 2 Small |
| Charxiv RQ · original / with Python | 85.6% | 77% | Inkling 2 Small |
| Chartography · original / with Python | 36.7% | 17.9% | Inkling 2 Small |
| STEM Reasoning | |||
| CritPt v2 · Full / Hard | 55.7% | 50.6% | Inkling 2 Small |
| HLE-Diamond | 28.7% | 25.3% | Inkling 2 Small |
| HLE (text only) | 44.1% | 43.3% | Inkling 2 Small |
| GPQA Diamond | 93% | 92.2% | Inkling 2 Small |
| Long Context | |||
| LongBench v2 | 70.6% | 69% | Inkling 2 Small |
| Factuality | |||
| AA-Omniscience v1.1 · score | 16.1 | 0.9 | Inkling 2 Small |
| AA-Omniscience v1.1 · non-hallucination rate | 73.3% | 35.9% | Inkling 2 Small |
| Instruction Following | |||
| ComplexConstraints v1.0 | 44.5% | 50.4% | DeepSeek-V4.1-Flash-1017 |
| Safety | |||
| FORTRESS (adversarial) | 83.8% | 72.8% | Inkling 2 Small |
| FORTRESS (benign) | 79.7% | 93.1% | DeepSeek-V4.1-Flash-1017 |
| StrongREJECT (none) | 97.1% | 99.4% | DeepSeek-V4.1-Flash-1017 |
| Cybersecurity | |||
| PatchEval | 79.6% | 80.4% | DeepSeek-V4.1-Flash-1017 |
Reading the comparison honestly
These numbers come from one table published by Thinking Machines, using third-party evaluations where those exist. That makes them comparable to each other, but it does not make them neutral: the vendor chose which benchmarks to run and how to present them, and a benchmark that a model loses has a way of not appearing.
The models are also not the same shape of thing. DeepSeek-V4.1-Flash-1017 sits in the open weights, smaller class tier of the card's own grouping, while Inkling 2 Small is the subject of this site. A win against a smaller model is less meaningful than a win against a larger one, and the table does not adjust for that.