Die Top-Modelle jeder Kategorie — kompiliert aus den Boards, die wir selbst für Modellwahl und Architekturempfehlungen heranziehen.
AA Intelligence Index v4.3.2 · Index (höher = besser) · Artificial Analysis
| # | Modell | Score |
|---|---|---|
| 1 |  Claude Opus 5.5 (max)Anthropic | 58 |
| 2 | Claude Sonnet 5.5 (max)Anthropic | 56 |
| 3 | Claude Opus 5.5 (high)Anthropic | 55 |
| 4 | Claude Fable 5.1 (max)Anthropic | 54 |
| 5 |  GPT-6 Astra (max)OpenAI | 53 |
| 6 |  Gemini 4 Argon (high)Google | 52 |
| 7 | GPT-6.1 Sol (max)OpenAI | 51 |
| 8 | Claude Opus 5.5 (medium)Anthropic | 50 |
MTEB (Overall) · Score (höher = besser) · MTEB Leaderboard (Hugging Face)
| # | Modell | Score |
|---|---|---|
| 1 | Gemini Embedding 001Google | 68.3 |
| 2 |  Qwen3-Embedding-8BAlibabaopen | 67.9 |
| 3 |  NV-Embed-v2NVIDIAopen | 67.3 |
| 4 | Voyage-Large-2Voyage AI | 67.0 |
| 5 | Cohere Embed v4Cohere | 66.7 |
| 6 | text-embedding-3-largeOpenAI | 66.5 |
| 7 | BGE-Large-v2BAAIopen | 65.9 |
Terminal-Bench 4.0 · % gelöste Aufgaben · Terminal-Bench (via Artificial Analysis)
| # | Modell | Score |
|---|---|---|
| 1 | Claude Sonnet 5.5Anthropic | 70.6 |
| 2 | Claude Opus 5.5Anthropic | 66.4 |
| 3 | Claude Mythos 5.1Anthropic | 60.9 |
| 4 | GPT-6 AstraOpenAI | 58.2 |
| 5 | Claude Fable 5.1Anthropic | 57.9 |
| 6 | Gemini 4 ArgonGoogle | 57.4 |
| 7 | GPT-6.1 SolOpenAI | 56.1 |
| 8 | Claude Opus 5Anthropic | 53.9 |
| 9 | Claude Fable 5Anthropic | 44.5 |
| 10 | GPT-6 SolOpenAI | 43.9 |
| 11 | GLM-5.3Z AIopenbestes Open-Weight-Modell | 41.8 |
Requests · Anfragen (Verwendung, nicht Benchmark) · OpenRouter
| # | Modell | Requests |
|---|---|---|
| 1 | Jev 1.13typesafeopen+4 % | 981M |
| 2 | Span-01 Literespanopen+9 % | 9.51M |
| 3 | d1liquidopenneu | 3.53M |
| 4 | GPT-6 Luna Decisionsopenaineu | 3.53M |
| 5 |  Clef Flashcloudflareopenneu | 3.46M |
| 6 | Clefcloudflareopenneu | 2.7M |
| 7 | Kev 4Bjaredpalmeropen+305 % | 2.45M |
| 8 |  Decider V1 27Bperplexityneu | 1.95M |
AA-Image-T2I v2.0 (Arena-Elo) · Elo aus Blind-Votings (höher = besser) · Artificial Analysis
| # | Modell | Score |
|---|---|---|
| 1 | GPT Image 2.5 SunburstOpenAI | 1198 |
| 2 | GPT Image 2.5 FlareOpenAI | 1191 |
| 3 | GPT Image 2OpenAI | 1172 |
| 4 | Nano Banana 2.1Google | 1160 |
| 5 | Grok Imagine Image 2.0xAI | 1156 |
| 6 |  MAI-Image-2.6Microsoft | 1151 |
Speech Arena (Arena-Elo) · Elo aus Blind-Votings (höher = besser) · Artificial Analysis
| # | Modell | Score |
|---|---|---|
| 1 |  Eleven v4 TurboElevenLabs | 1332 |
| 2 | Eleven v4ElevenLabs | 1324 |
| 3 | Qwen-Audio-3.1-TTS-PlusAlibabaopen | 1295 |
| 4 | Sonic 3.6Cartesia | 1279 |
| 5 | Gemini 3.8 Flash TTSGoogle | 1276 |
| 6 | Qwen-Audio-3.0-TTS-PlusAlibabaopen | 1263 |
AA-WER v2 · Fehlerrate (niedriger = besser) · Artificial Analysis
| # | Modell | WER |
|---|---|---|
| 1 | StepAudio 3 ASRStepFunopen | 1.7 % |
| 2 | Fun-Realtime-ASR-previewAlibabaopen | 1.7 % |
| 3 | MAI-Transcribe-2Microsoft | 2.0 % |
| 4 | Scribe v2ElevenLabs | 2.2 % |
| 5 | Grok Voice Transcribe 2.0xAI | 2.3 % |
| 6 | Universal-3 ProAssemblyAI | 3.1 % |
| 7 | GPT TranscribeOpenAI | 3.3 % |