📚 Knowledge Q&A Assistant frontier
As of November 16, 2026, the highest Knowledge Q&A Assistant score on this frontier is Claude Fable 5.2 at 100.0, with an estimated monthly cost of $2,640.00.
An internal help desk or search assistant that answers questions from documents and general knowledge. Factual accuracy (SimpleQA) leads, with long-context retrieval for pasted documents, broad everyday ability, human preference for clear answers, and a slice of hard expert questions. Cost is weighted for steady daily volume.
The y axis is this scenario's weighted score; the x axis starts at estimated monthly cost for 2,000 requests a day. Switch axes to compare against speed, latency or context. See the full ranking.
On the frontier
| Model | Score | Est. $/month |
|---|---|---|
| Claude Fable 5.2Anthropic | 100.0 | $2,640 |
| Muse WonderMeta | 89.9 | $1,320 |
| Muse Spark 1.3Meta | 89.8 | $353 |
| Gemini 4 CarbonGoogle | 87.0 | $330 |
| Gemini 2.5 ProGoogle | 78.7 | $263 |
| DeepSeek R1DeepSeek | 78.1 | $155 |
| Qwen3 235BAlibaba (Qwen) | 77.1 | $49 |
| DeepSeek V4.1 FlashDeepSeek | 63.6 | $38 |
| Gemini 1.5 FlashGoogle | 62.0 | $23 |
| GPT-6 LunaOpenAI | 46.2 | $17 |
| DeepSeek V4 Flash 0731DeepSeek | 40.5 | $16 |
| GPT-5 nanoOpenAI | 17.1 | $11 |