Models
One row per model line, latest only. Intelligence and speed from Artificial Analysis; prices from vendor pages. As of July 18, 2026.
| Frontier · 7 | |
| GPT-5.6 Luna gpt-5.6-luna | beats 11 models |
| Muse Spark 1.1 muse-spark-1.1 | beats 9 models |
| DeepSeek V4 Flash deepseek-v4-flash | beats 5 models |
| GLM-5.2 glm-5.2 | beats 4 models |
| GPT-5.4 Nano gpt-5.4-nano | beats 2 models |
| Gemini 3.1 Flash-Lite gemini-3.1-flash-lite | beats 2 models |
| Kimi K3 kimi-k3 | beats 2 models |
| One axis · 12 | |
| GPT-5.6 Sol vs Gemini 3.5 Flash | holds only intel — 59 vs 50 |
| GPT-5.6 Terra vs Gemini 3.5 Flash | holds only intel — 55 vs 50 |
| GPT-5.4 Mini vs Gemini 3.1 Flash-Lite | holds only intel — 40 vs 25 |
| Claude Fable 5 vs Gemini 3.5 Flash | holds only intel — 60 vs 50 |
| Claude Sonnet 5 vs Gemini 3.5 Flash | holds only intel — 53 vs 50 |
| Claude Haiku 4.5 vs DeepSeek V4 Flash | holds only inputs — vision vs text |
| Gemini 3.5 Flash vs GLM-5.2 | holds only inputs — omni vs text |
| Grok 4.5 vs DeepSeek V4 Flash | holds only intel — 54 vs 40 |
| Grok 4.3 vs GPT-5.6 Luna | holds only output price — $2.5 vs $6 |
| DeepSeek V4 Pro vs DeepSeek V4 Flash | holds only intel — 44 vs 40 |
| Qwen3.7 Max vs GPT-5.6 Luna | holds only tok/s — 205 vs 197 |
| MiniMax M3 vs DeepSeek V4 Flash | holds only intel — 44 vs 40 |
| Dominated · 2 | |
| Claude Opus 4.8 vs Kimi K3 | worse or equal on every axis |
| Gemini 3.1 Pro vs Gemini 3.5 Flash | worse or equal on every axis |
| Unscored · 1 | |
| GPT-5.5 Pro | no AA score — can't be placed |
Intel / tok/s: Artificial Analysis Intelligence Index, at each model's top reasoning effort. A beats B when B holds at most one axis of intel, tok/s, ctx, prices, inputs, weights. Recomputed from the data on every load. Not a list of what CrabTalk runs.