OUR PROPRIETARY RESEARCH BENCHMARK
REVAL™
research evaluation.
Research Evaluation Index measures real-world AI research on a 0–100 scale. Scores are published only after actual evaluations and independent grading.
Methodology & weights →| Model | REVAL™ |
|---|---|
| GPT-6 Astra | Not tested |
| GPT-6 Sol | Not tested |
| GPT-6 Luna | Not tested |
| GPT-6.1 Sol | Not tested |
| GPT-5.6 Sol | Not tested |
| GPT-5.6 Terra | Not tested |
| GPT-5.6 Luna | Not tested |
| Claude Fable 5.1 | Not tested |
| Claude Opus 5.5 | Not tested |
| Claude Sonnet 5.5 | Not tested |
| Claude Haiku 4.5 | Not tested |
| Gemini 3.8 Flash | Not tested |
| Gemini 3.7 Flash | Not tested |
| Gemini 3.6 Flash | Not tested |
| Gemini 3.5 Flash | Not tested |
| Gemini 3.5 Flash-Lite | Not tested |
| Gemini 3.1 Pro Preview | Not tested |
| Grok 4.6 | Not tested |
| Grok 4.5 | Not tested |
| DeepSeek V4.1 Flash | Not tested |
| DeepSeek V4 Pro | Not tested |
| Mistral Medium 3.5 | Not tested |
| Mistral Small 4 | Not tested |
| Mistral Large 3 | Not tested |
| Llama 4 Scout | Not tested |
| Llama 4 Maverick | Not tested |
| Qwen3.5-397B-A17B | Not tested |
| Qwen3.5-122B-A10B | Not tested |
| gpt-4.1 | Not tested |
| gpt-4.1-mini | Not tested |
| gpt-4.1-nano | Not tested |
| gpt-5 | Not tested |
| gpt-5-mini | Not tested |
| gpt-5-nano | Not tested |
| gpt-5.1 | Not tested |
| gpt-5.2 | Not tested |
| gpt-5.4-mini | Not tested |
| gpt-5.4-nano | Not tested |
| Gemini 4 Argon | Not tested |
| Muse Wonder | Not tested |
| Muse Wonder Full | Not tested |