German LLM Index - SB10K Sentiment: leaderboard
German Language LLM Index (dach.peerbench.ai, a PeerBench project, 2026), which reruns every model itself with lm-evaluation-harness: chat template, temperature 0, up to 8,192 generated tokens, thinking models set to minimal reasoning effort where the API allows it and other models run as served. Labels carry (Reasoning off) when the site classed the run as reasoning-off (at most ten reasoning tokens, or under 0.1% of its completion tokens); other rows reasoned at the provider default or at the effort the label names. Native German tweets (SB10K) labelled positive, neutral or negative; the model names the label for 1,024 tweets and the score is the Matthews correlation coefficient times 100 (0 is chance level, 100 perfect, negative values worse than chance). The tweets come from the EuroEval SB10K split, but the run uses lm-evaluation-harness with the prompt of the site rather than the EuroEval harness, so it is kept apart from the EuroEval German SB10K board. Higher is better. Source: https://dach.peerbench.ai/
Metric: Matthews correlation coefficient (x100). Source: dach.peerbench.ai. Saturation forecast: Around February 2027. 44 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 70.15 | #64 |
| 2 | Gemini 3.7 Flash | 67.53 | #24 |
| 3 | Claude Fable 5 | 67.33 | #20 |
| 4 | Gemini 3.8 Flash | 66.92 | #30 |
| 5 | Muse Spark Open | 66.41 | #42 |
| 6 | Claude Opus 4.8 (Non-reasoning) | 64.52 | #39 (Claude Opus 4.8) |
| 7 | Gemini 3 Flash (Preview) (Non-reasoning) | 63.57 | #90 (Gemini 3 Flash (Preview)) |
| 8 | GLM-5.3 Flash | 63.56 | #77 |
| 9 | Qwen 3.7 Max (Non-reasoning) | 63.17 | #85 (Qwen 3.7 Max) |
| 10 | GPT-5.6 Terra (Non-reasoning) | 62.74 | #40 (GPT-5.6 Terra) |
| 11 | GPT-5.5 (Low) | 62.45 | #33 (GPT-5.5) |
| 12 | DeepSeek V4 Flash (Non-reasoning) | 62.38 | #155 (DeepSeek V4 Flash) |
| 13 | GPT-5.6 Luna (Non-reasoning) | 61.88 | #79 (GPT-5.6 Luna) |
| 14 | MiMo-V2.5-Pro (Non-reasoning) | 61.69 | #167 (MiMo-V2.5-Pro) |
| 15 | Gemini 2.5 Flash (Non-reasoning) | 61.57 | #255 (Gemini 2.5 Flash) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=german-llm-index-sb10k-sentiment · How It Works · Data refreshed daily, snapshot 2026-11-05.