Which AI can you trust for baseline knowledge?
By Anthony Vincent Caracappa, Founder and Self-Directed Space Researcher, Viventine Space Systems · updated
Space history and orbital mechanics: facts and calculations that do not change.
On their own: Claude Fable 5.1, Claude Haiku 5.5, Claude Opus 5.5, Claude Sonnet 5.5, Grok 4.7, gpt-oss 20B. With Orbit Sentinel tools: Claude Fable 5.1, Claude Haiku 5.5, Claude Opus 5.5, Claude Sonnet 5.5, Grok 4.7, Gemma 4 12B.
| AI | AloneClosed-book | With dataWith Orbit Sentinel tools | n | ||||
|---|---|---|---|---|---|---|---|
| right / wrong | range | declined | right / wrong | range | declined | ||
| Claude Fable 5.1 Anthropic | 100 / 0 | 87–100 | 0 | 100 / 0 | 87–100 | 0 | 25×3 |
| Claude Haiku 5.5 Anthropic | 100 / 0 | 87–100 | 0 | 97 / 0 | 82–100 | 3 | 25×3 |
| Claude Opus 5.5 Anthropic | 100 / 0 | 87–100 | 0 | 100 / 0 | 87–100 | 0 | 25×3 |
| Claude Sonnet 5.5 Anthropic | 100 / 0 | 87–100 | 0 | 99 / 1 | 84–100 | 0 | 25×3 |
| Grok 4.7 xAI | 96 / 0 | 80–99 | 4 | 97 / 0 | 82–100 | 3 | 25×3 |
| gpt-oss 20B OpenAI | 96 / 4 | 80–99 | 0 | 85 / 5 | 66–94 | 10 | 25×3 |
| Gemma 4 12B Google | 85 / 15 | 58–96 | 0 | 85 / 1 | 66–94 | 13 | 24×3 |
| DeepSeek-R1 14B DeepSeek | 68 / 32 | 47–84 | 0 | — | 22×1 | ||
% right / % wrong per cell; color is the verdict (green trust it, amber verify, red don't use). Range is the 95% interval for % right, counting each question once; n is questions × runs for the tools track.