KI-Modell-Rangliste — Globales Ranking 2026
330 Live benchmark rankings across 330+ AI models by Intelligence Index, SWE-bench Verified, GPQA Diamond, speed, and API token cost.
Intelligence Index vs. Blended API Price (Pareto Frontier)
Emerald-ringed nodes represent Pareto-optimal frontier models: no competing model offers higher intelligence at a lower 3:1 blended token cost.
Gemini 4 Argon
Gemini 4 Argon von Google ist ein leistungsstarkes KI-Modell mit einem Kontextfenster von 1M Token zu API-Preisen von $2/1M Input- und $10/1
| # | Model | providers | Intelligence ▼ | Arena Elo | SWE-bench | GPQA | Speed | Cost / task (3:1) | Pricing | Context |
|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Gemini 4 ArgonPareto | 97.0 | 1525 | 81.9% | 81.5% | 115 | $4 | $2 / $10 | 1M | |
| 2 | GPT-6 Astra | OpenAI | 96.0 | 1520 | 82.4% | 80.9% | 100 | $20 | $10 / $50 | 272K |
| 3 | Claude Opus 5.5 | Anthropic | 94.0 | 1507 | 84.6% | 82.3% | 78 | $8 | $4 / $20 | 1M |
| 4 | GPT-6.1 Sol | OpenAI | 94.0 | 1519 | 82.3% | 80.8% | 120 | $4 | $2 / $10 | 1M |
| 5 | Llama 5OpenPareto | Meta | 94.0 | 1485 | 74% | 84% | 65 | Free | Free | 5M |
| 6 | GPT-6 Sol | OpenAI | 93.0 | 1516 | 82.1% | 80.6% | 116 | $4 | $2 / $10 | 1M |
| 7 | Kimi K3Open | Moonshot AI | 93.0 | 1480 | 74.5% | 79% | 58 | Free | Free | 1.0M |
| 8 | Llama 4 BehemothOpen | Meta | 93.0 | 1490 | 75% | 82% | 45 | Free | Free | 10M |
| 9 | Qwen3.8 Max PreviewOpen | Alibaba | 93.0 | 1485 | 75.2% | 79.5% | 60 | Free | Free | 1M |
| 10 | Grok-4 | xAI | 92.0 | 1507 | 76.5% | 77.8% | 125 | $4.38 | $2.5 / $10 | 1M |
| 11 | Qwen3.7 Max | Alibaba | 92.0 | 1470 | 73.5% | 78.5% | 80 | $3.15 | $1.8 / $7.2 | 1M |
| 12 | Claude Sonnet 4.5 | Anthropic | 91.0 | 1355 | 77.8% | 70.2% | 90 | $6 | $3 / $15 | 1M |
| 13 | Gemini Deep Research | 91.0 | 1495 | — | 85% | 60 | $17.50 | $10 / $40 | 1.0M | |
| 14 | GLM-5.2Open | Zhipu AI | 91.0 | 1440 | 67.5% | 73.5% | 60 | Free | Free | 1M |
| 15 | ERNIE 5.0 | Baidu | 90.0 | 1430 | 60% | 74.5% | 70 | $2.63 | $1.5 / $6 | 128K |
| 16 | Muse Spark 1.1 | Meta | 90.0 | 1495 | 76.5% | 83.5% | 105 | $6 | $3 / $15 | 1M |
| 17 | Qwen3.6 Max Preview | Alibaba | 90.0 | 1445 | 70% | 76% | 75 | $2.63 | $1.5 / $6 | 262K |
| 18 | GLM-5.1Open | Zhipu AI | 89.0 | 1410 | 63% | 70% | 65 | Free | Free | 200K |
| 19 | GPT-5.6 Sol | OpenAI | 89.0 | 1496 | 83.8% | 82% | 108 | $8 | $4 / $20 | 1.1M |
| 20 | Kimi K2.7 CodeOpen | Moonshot AI | 89.0 | 1415 | 62% | 73% | 75 | Free | Free | 262K |
| 21 | MiniMax-M3Open | MiniMax | 89.0 | 1420 | 65% | 72% | 70 | Free | Free | 524K |
| 22 | Qwen3.7 Plus | Alibaba | 89.0 | 1435 | 68% | 75% | 92 | $1.40 | $0.8 / $3.2 | 1M |
| 23 | Claude Mythos 5.1 | Anthropic | 88.0 | 1520 | 84.2% | 82% | 60 | $30 | $15 / $75 | 1M |
| 24 | Doubao Pro | ByteDance | 88.0 | 1410 | 58% | 71.5% | 85 | $1.40 | $0.8 / $3.2 | 256K |
| 25 | Nano Banana 2 | 88.0 | 1455 | — | — | 65 | $9.75 | $3 / $30 | 131K | |
| 26 | Qwen3.6 Plus | Alibaba | 88.0 | 1425 | 66.5% | 73.8% | 90 | $1.23 | $0.7 / $2.8 | 1M |
| 27 | Qwen3 Max | Alibaba | 88.0 | 1395 | 62.5% | 71.2% | 72 | $3 | $2 / $6 | 262K |
| 28 | Veo 3.1 | 88.0 | 1450 | — | — | 14 | $117.75 | $7 / $450 | 4K | |
| 29 | Gemini 3.8 Flash | 87.0 | 1480 | 74.8% | 84% | 175 | $1.50 | $0.75 / $3.75 | 1.0M | |
| 30 | GLM-5Open | Zhipu AI | 87.0 | 1390 | 58% | 68% | 68 | Free | Free | 205K |
Methodology: Intelligence score is calculated from verified composite benchmarks (SWE-bench Verified, GPQA Diamond, HumanEval, MMLU-Pro). Prices are per 1M tokens from official provider documentation.
AI Benchmark, Coding & Speed Leaderboard FAQ
Häufig gestellte Fragen (FAQ)
How to use Claude 3.7 Sonnet for full-stack web development?+
Claude 3.7 Sonnet excels at full-stack development by combining architectural refactoring with extended thinking mode, achieving a 70.3% verified score on SWE-bench for autonomous multi-file bug fixing.
Is Claude 3.7 Sonnet better than GPT-4o for complex Python refactoring?+
Yes, Claude 3.7 Sonnet holds a decisive lead on multi-module Python refactoring due to its configurable extended thinking budget and state-of-the-art SWE-bench scores compared to GPT-4o standard generation.
Can DeepSeek R1 replace Claude 3.7 Sonnet for software engineering?+
DeepSeek R1 matches frontier models on algorithmic and mathematical logic at 10x lower API costs, but Claude 3.7 Sonnet retains the overall lead in end-to-end repository navigation and agentic tool use.
What is the best AI model for automated unit test generation in 2026?+
Claude 3.7 Sonnet and OpenAI o3-mini lead in automated unit test generation, delivering high branch coverage and minimizing edge-case regressions through rigorous chain-of-thought verification.
What is the SWE-bench verified score difference between Claude 3.7 Sonnet and o3-mini?+
Claude 3.7 Sonnet scores 70.3% on SWE-bench Verified in extended thinking mode, whereas OpenAI o3-mini scores approximately 49%, reflecting Claude specialized architecture for production software maintenance.
What is the difference between DeepSeek R1 reasoning tokens and standard output tokens?+
Reasoning tokens represent the internal chain-of-thought scratchpad generated during thinking mode. In DeepSeek R1, both reasoning tokens and visible output tokens are billed at the same economical $2.19/1M rate.
How does Claude 3.7 Sonnet hybrid reasoning mode work?+
Claude 3.7 Sonnet combines instant response inference with optional extended thinking mode, allowing developers to set an exact reasoning token budget between 1,024 and 128,000 tokens per API call.
Which model is better for competitive math and programming: DeepSeek R1 or o3-mini?+
Both models achieve over 90% on MATH 500; DeepSeek R1 excels on raw mathematical proofs, while o3-mini provides tighter algorithmic code generation with lower token overhead.