A-Fantasia Spelling (Valid Attempts with Retries)
Games · 2026-09-30
A-Fantasia backward-spelling error among valid responses under the retry protocol. Infers a constrained WordNet noun from its definition and reverses it. Latest run with at least 80 valid items; up to five retries for unscorable responses; denominator is not published.
Top models (lower is better)
| Model | Score |
|---|---|
| Claude Opus 5 | 15.0 |
| Opus 4.1 | 15.0 |
| Opus 4.5 | 16.0 |
| Opus 4.6 | 16.2 |
| Opus 4.8 | 17.0 |
| Opus 4 | 22.0 |
| GPT-5.6 Sol | 23.0 |
| Sonnet 4.5 | 24.0 |
| GPT-5.4 | 26.0 |
| GPT-5.5 | 27.0 |
| GPT-4.5 | 32.0 |
| Sonnet 4 | 32.0 |
| GPT-5.6 Terra | 33.0 |
| Claude 3.5 Sonnet (Oct 2024) | 34.0 |
| Claude 3.7 Sonnet | 35.0 |
| Opus 3 | 35.0 |
| Grok 3 (API) | 36.4 |
| GPT-5.2 | 37.0 |
| Sonnet 4.6 | 40.0 |
| DeepSeek-V4-Pro | 45.0 |
| Qwen3.7-Max (source unspecified) | 46.0 |
| GPT-4o (source unspecified) | 56.6 |
| GPT-5 Chat | 62.0 |
| Inkling | 62.0 |
| GPT-5.1 | 63.0 |
| GPT-4.1 | 64.0 |
| Gemini 2.5 Flash | 64.0 |
| Haiku 4.5 | 65.0 |
| DeepSeek V4.1 Flash | 65.0 |
| Mistral Large 4 Preview | 66.0 |
| Claude Haiku 5.5 | 70.0 |
| GPT-5.6 Luna | 71.0 |
| Kimi K2.6 | 72.0 |
| Kimi K2 Instruct | 72.0 |
| Qwen3.6 Plus (2026-04-02) | 72.0 |
| Hy4 Preview | 73.0 |
| Qwen3.7-Plus (source unspecified) | 74.0 |
| GLM-5.2 | 76.3 |
| GLM-5.1 | 76.8 |
| Gemini 2.0 Flash | 77.0 |