AI-modeller
Rangering fra Arena.ai, blindtester og agentoppgaver.
Agentiske oppgaver med flerstegsresonnering og verktøybruk.
| # | Modell | Leverandør | Netto forbedring |
|---|---|---|---|
| 01 | Claude Opus 5 (High) | Anthropic | +12,47 % |
| 02 | Claude Opus 5 (Max) | Anthropic | +12,00 % |
| 03 | Claude Fable 5 (High) | Anthropic | +11,57 % |
| 04 | Kimi K3 (Max) | Moonshot | +10,41 % |
| 05 | GPT 5.6 Sol (xHigh) | OpenAI | +9,74 % |
| 06 | Claude Opus 4.8 (High) | Anthropic | +9,55 % |
| 07 | GPT 5.5 (xHigh) | OpenAI | +8,51 % |
| 08 | Claude Opus 4.7 (High) | Anthropic | +8,12 % |
| 09 | GPT 5.5 (High) | OpenAI | +7,74 % |
| 10 | Claude Opus 4.7 | Anthropic | +7,40 % |
| 11 | Claude Sonnet 5 (High) | Anthropic | +6,62 % |
| 12 | Claude Opus 4.6 | Anthropic | +6,60 % |
| 13 | GPT 5.5 | OpenAI | +6,38 % |
| 14 | DeepSeek V4 Pro (High) (0813) | DeepSeek | +6,26 % |
| 15 | Qwen3.8 Max | Alibaba | +6,20 % |
| 16 | Grok 4.5 | SpaceXAI | +6,17 % |
| 17 | GLM 5.2 (Max) | Z.ai | +5,82 % |
| 18 | GPT 5.4 (High) | OpenAI | +4,98 % |
| 19 | GPT 5.6 Luna (xHigh) | OpenAI | +4,04 % |
| 20 | Deepseek V4 Flash (High) (20260731) | DeepSeek | +3,99 % |
| 21 | Gemini 3.7 Flash (High) | +3,32 % | |
| 22 | GPT 5.6 Terra (xHigh) | OpenAI | +3,19 % |
| 23 | Claude Sonnet 4.6 | Anthropic | +2,88 % |
| 24 | Claude Opus 4.8 | Anthropic | +2,51 % |
| 25 | Muse Spark 1.2 (xHigh) | Meta | +2,06 % |
Data fra arena.ai. Agent-boardet oppdateres daglig, sist 23. august 2026. Kode-boardet er snapshot 16. juni 2026.