# Same Performance. Full Control. GDPval-AA v2 leaderboard · Elo GDPval-AA v2 evaluates models on real-world work tasks and expresses performance as an Elo rating, anchored to a human baseline of 1,000 — higher is better. Human baseline (1,000) 1. Claude Fable 5.1 1764 Claude Fable 5.1 Anthropic Elo rating: 1764 Released: August 2026 Anthropic’s currently strongest closed model — leads the ranking. 2. Claude Opus 5 1735 Claude Opus 5 Anthropic Elo rating: 1735 Released: June 2026 Anthropic’s new Opus flagship — a close second behind Fable 5.1. 3. GLM 5.3 on Platte 1676 GLM 5.3 on Platte Z.ai Elo rating: 1676 Released: August 2026 Z.ai’s open flagship — ranks above GPT-6, and therefore above every OpenAI model. Runs on Platte’s own infrastructure in Europe, under European law. 4. Qwen 3.8 on Platte 1628 Qwen 3.8 on Platte Alibaba Elo rating: 1628 Released: August 2026 Alibaba’s open flagship — also still above GPT-6. Served on Platte in Europe, under European law. 5. GPT-6 1580 GPT-6 OpenAI Elo rating: 1580 Released: August 2026 OpenAI’s frontier model — ranked below the open models GLM 5.3 and Qwen 3.8. 6. Gemini 3.8 Flash (high) 1464 Gemini 3.8 Flash (high) Google Elo rating: 1464 Released: August 2026 Google’s efficient Flash model in its high-reasoning mode — clearly above the human baseline, but behind the frontier. 7. Mistral Medium 3.5 875 Mistral Medium 3.5 Mistral Elo rating: 875 Released: November 2025 The mid-tier offering of Europe’s largest model maker — the only model shown below the human baseline of 1,000. [Chart: artificialanalysis.ai](https://artificialanalysis.ai/) · As of: September 4, 2026 · All models shown are the current top models of their respective companies. Models on Platte ## Current model lineup, on EU infrastructure Every model runs on Platte in European data centres, under European law. Intelligence and cost per task are measured by Artificial Analysis, independently of the model makers. ### GLM 5.3 Z.ai Intelligence 45 Avg. cost per task $2.01 Input / output €2.00 / €5.20 per 1M tokens Context window 1M tokens - Reasoning ### GLM 5.3 Flash Z.ai Intelligence 42 Avg. cost per task $0.28 Input / output €0.60 / €1.20 per 1M tokens Context window 1M tokens - Reasoning - Image Vision ### Qwen 3.8 2.4T (A95B) Alibaba Intelligence 40 Avg. cost per task $2.16 Input / output €3.00 / €6.50 per 1M tokens Context window 262K tokens - Reasoning ### Qwen 3.8 Flash Next Alibaba Intelligence 40 Avg. cost per task $0.37 Input / output €0.60 / €1.20 per 1M tokens Context window 256K tokens - Reasoning - Image Vision ### DeepSeek V4.1 Flash DeepSeek Intelligence 39 Avg. cost per task $0.27 Input / output €0.75 / €2.50 per 1M tokens Context window 1M tokens - Reasoning - Image Vision ### DeepSeek V4 Flash (0731) DeepSeek Intelligence 34 Avg. cost per task $0.22 Input / output €0.60 / €0.90 per 1M tokens Context window 1M tokens - Reasoning Intelligence, cost per task, context window and capabilities: [Artificial Analysis](https://artificialanalysis.ai/leaderboards/models) (Intelligence Index v4.3.2, each at maximum reasoning effort), as of October 14, 2026. Artificial Analysis computes the cost per task at its own reference prices in US dollars — it compares the models with each other and is not the price on Platte. Prices for the API and extra usage: [Platte price list](/documents/26-09-21_Preisliste-Platte-AP.pdf), as of September 21, 2026, excl. VAT.