Up to four models
Claude Opus 5 vs Inkling-Small
On the index (GPT-4 = 100), Claude Opus 5 scores 185 and Inkling-Small 153. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.
Claude Opus 5
185169–201
WeightsClosed
ReleasedJul 2026
Price in / out$5.0 / $25
Context1M
Core tests taken6 of 9
TMInkling-Small
153137–169
WeightsOpen
ReleasedJul 2026
Price in / out$0.45 / $1.2
Context524K
Core tests taken5 of 9
Claude Opus 5 leads Inkling-Small by 32 points, ahead on 5 of 5 shared core tests, at 16× the price.
Test by testthird-party results
Claude Opus 5TMInkling-Small
Every result as a table
| Test | Claude Opus 5 | Inkling-Small |
|---|---|---|
| Reasoning & knowledge | ||
| GPQA Diamond (Epoch AI) | 93.9% | 88.5% |
| GPQA Diamond (Vals AI) | 93.4% | 83.6% |
| MMLU Pro | 91.6% | 85.6% |
| SimpleQA Verified | 59.9% | 19.1% |
| Maths | ||
| FrontierMath (tiers 1-3) | 85.6% | 46.3% |
| AIME-style maths (OTIS mock) | 98.9% | 90.0% |
| ProofBench v1.1 | 99.0% | 6.0% |
| FrontierMath (tier 4) | 73.2% | 17.1% |
| Coding | ||
| Terminal-Bench 4.0 | 53.5% | 0.0% |
| Code Migration | 57.5% | 13.7% |
| CyberBench v1.1 | 65.4% | 64.1% |
| IOI | 84.3% | 9.3% |
| LiveCodeBench | 89.0% | 85.9% |
| ProgramBench | 3.0% | 0.5% |
| SWE-bench | 97.0% | 82.2% |
| Vibe Code Bench v1.1 | 88.4% | 19.1% |
| Agents | ||
| SkillsBench | 60.4% | 33.6% |
| Mystery game puzzles | 59.0% | 6.0% |
| Professional work | ||
| Finance Agent (v2) | 58.6% | 41.3% |
| CorpFin v2 | 73.2% | 69.6% |
| EMB | 73.6% | 31.8% |
| LegalBench | 87.0% | 83.0% |
| Legal Research Bench | 55.3% | 25.5% |
| MedCode | 63.6% | 37.9% |
| MedScribe | 91.0% | 84.1% |
| MortgageTax | 72.1% | 62.3% |
| Public Benefits Bench v1.1 | 76.9% | 59.4% |
| Tax Agent Bench | 47.1% | 15.4% |
| TaxEval v2 | 75.1% | 75.5% |
| Novel problems | ||
| ARC-AGI-2 | 90.4% | 40.1% |
| ARC-AGI-1 | 97.5% | 84.0% |
Tests at least two of these models have taken. Bold rows are the core tests.