Inkling 2 Small vs Claude Haiku 5.5
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 Claude Haiku 5.5 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 Code2Video-AA.
Head to head
- Position
- Inkling 2 Small: the subject of this site · Claude Haiku 5.5: closed weights
- Bechmarks scored
- 48 shared rows, out of 71 total
- Record
- 27 wins · 21 losses · 0 ties
Largest differences
| Benchmark | Inkling 2 Small | Claude Haiku 5.5 | Gap |
|---|---|---|---|
| Code2Video-AA · Visual Creation | 1,020 | 1,192 | -172 |
| DesignArena Game Dev · Visual Creation | 1,428 | 1,319 | + 109 |
| DesignArena Website · Visual Creation | 1,340 | 1,261 | + 79 |
| DesignArena 3D Design · Visual Creation | 1,401 | 1,334 | + 67 |
| Code Arena · Visual Creation | 1,653 | 1,589 | + 64 |
| AA-Briefcase v1.1 · Knowledge Work | 1,527 | 1,574 | -47 |
| SWEBench Pro v2 (Hard) · Coding | 86.3% | 52.9% | + 33.4 |
| JobBench · Knowledge Work | 70.5% | 39.2% | + 31.3 |
| AA-3D · Visual Creation | 1,021 | 994 | + 27 |
| Terminal Bench v4.0 · Coding | 13.1% | 32.8% | -19.7 |
Every shared benchmark
| Benchmark | Inkling 2 Small | Claude Haiku 5.5 | Leader |
|---|---|---|---|
| Third Party Index | |||
| AA Index (v5.0) | 30.6 | 36.5 | Claude Haiku 5.5 |
| Vals Index (v2.2) | 56.1% | 58.7% | Claude Haiku 5.5 |
| Coding | |||
| FrontierCode v2.0 | 25.4% | 34.9% | Claude Haiku 5.5 |
| DeepSWE v1.1 | 72.1% | 59.9% | Inkling 2 Small |
| Terminal Bench Science v0.1 | 5.7% | 20% | Claude Haiku 5.5 |
| Terminal Bench v4.0 | 13.1% | 32.8% | Claude Haiku 5.5 |
| Vibe Code Bench v1.1 | 81.6% | 90.4% | Claude Haiku 5.5 |
| SWE-Marathon v1.1 | 38.8% | 50% | Claude Haiku 5.5 |
| SWEBench Multilingual | 81.2% | 83.7% | Claude Haiku 5.5 |
| SWEBench Multimodal v2 | 43.1% | 44% | Claude Haiku 5.5 |
| SWEBench Pro v2 (Hard) | 86.3% | 52.9% | Inkling 2 Small |
| SWEBench Pro v2 (Full) | 97.7% | 87.2% | Inkling 2 Small |
| SWEBench Verified | 82.2% | 88.9% | Claude Haiku 5.5 |
| Autoresearch | |||
| BioMysteryBench | 70% | 70.4% | Claude Haiku 5.5 |
| Search and Tools | |||
| MCP Atlas · Full / Public | 85.4% | 81% | Inkling 2 Small |
| Knowledge Work | |||
| AA-Briefcase v1.1 | 1,527 | 1,574 | Claude Haiku 5.5 |
| GDPVal-AA v2.1 | 1,630 | 1,631 | Claude Haiku 5.5 |
| JobBench | 70.5% | 39.2% | Inkling 2 Small |
| Enterprise Operations | |||
| AutomationBench v1.0.6 | 45.6% | 35.4% | Inkling 2 Small |
| Finance Work | |||
| Finance Agent v2.1 | 67.5% | 65.4% | Inkling 2 Small |
| Excel Modeling Benchmark v1.1 | 69.5% | 65.5% | Inkling 2 Small |
| Tax Agent Bench | 65.8% | 62.2% | Inkling 2 Small |
| Apex-Accounting v1.1 · pass@1 / mean score | 14.5% | 15.3% | Claude Haiku 5.5 |
| Diligence Bench · Best harness | 58.1% | 40.8% | Inkling 2 Small |
| Legal Work | |||
| Harvey LAB v1.1 | 21.7% | 4.2% | Inkling 2 Small |
| Legal Research Bench | 46.6% | 43.8% | Inkling 2 Small |
| Visual Creation | |||
| DesignArena Website | 1,340 | 1,261 | Inkling 2 Small |
| DesignArena Game Dev | 1,428 | 1,319 | Inkling 2 Small |
| DesignArena 3D Design | 1,401 | 1,334 | Inkling 2 Small |
| Code Arena | 1,653 | 1,589 | Inkling 2 Small |
| AA-3D | 1,021 | 994 | Inkling 2 Small |
| Code2Video-AA | 1,020 | 1,192 | Claude Haiku 5.5 |
| Visual Reasoning | |||
| GDP.pdf | 25.2% | 20.8% | Inkling 2 Small |
| MMMU Pro (Standard 10) | 81.6% | 79.1% | Inkling 2 Small |
| Charxiv RQ · original / with Python | 85.6% | 84.4% | Inkling 2 Small |
| Chartography · original / with Python | 36.7% | 46.4% | Claude Haiku 5.5 |
| SAGE v1.1 | 88.6% | 89.8% | Claude Haiku 5.5 |
| STEM Reasoning | |||
| CritPt v2 · Full / Hard | 55.7% | 64% | Claude Haiku 5.5 |
| HLE-Diamond | 28.7% | 29.6% | Claude Haiku 5.5 |
| HLE (text only) | 44.1% | 44.4% | Claude Haiku 5.5 |
| GPQA Diamond | 93% | 89.8% | Inkling 2 Small |
| Long Context | |||
| AA-LCR v1.1 | 83.3% | 82.7% | Inkling 2 Small |
| LongBench v2 | 70.6% | 65.3% | Inkling 2 Small |
| Factuality | |||
| AA-Omniscience v1.1 · score | 16.1 | 12.3 | Inkling 2 Small |
| AA-Omniscience v1.1 · non-hallucination rate | 73.3% | 65.8% | Inkling 2 Small |
| Instruction Following | |||
| ComplexConstraints v1.0 | 44.5% | 36.8% | Inkling 2 Small |
| Safety | |||
| DistressBench | 82% | 71.7% | Inkling 2 Small |
| Cybersecurity | |||
| CyberBench v1.1 | 72.7% | 75.5% | Claude Haiku 5.5 |
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. Claude Haiku 5.5 sits in the closed weights 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.