Technical specifications, benchmark results, and pricing from frontier AI labs worldwide — with verified sources and effective dates.
Full telemetry: parameters, context windows, benchmark scores, and API pricing per 1M tokens.
| Model & Datasheet | Lab / Creator | Region | Architecture / Params | Context | MMLU-Pro | SWE-bench | MATH | Pricing (Input / Output) | Effective Date |
|---|---|---|---|---|---|---|---|---|---|
| China | 238B (21B active) Multi-head Latent Attention (MLA) + DeepSeekMoE | 256k tokens | 90.6% | 61.2% | 93.1% | $0.07 / 1M tokens $0.14 / 1M tokens | 31.07.2026 | ||
| United States | 70B Dense Dense Native-Multimodal Open Weights | 256k tokens | 91.5% | 63.4% | 93.8% | Free / Self-Hosted Free / Self-Hosted | 25.08.2026 | ||
| United States | 14B Dense Compact High-Efficiency Open Weights | 128k tokens | 86.2% | 49.0% | 86.5% | Free / Self-Hosted Free / Self-Hosted | 25.08.2026 | ||
| United States | 30B (6B active) Multimodal Agentic Sparse MoE | 128k tokens | 87.9% | 54.1% | 88.2% | Free / Self-Hosted Free / Self-Hosted | 10.09.2026 | ||
| United States | 3.5B Dense Ultra-Compact Edge Intelligence Transformer | 64k tokens | 79.4% | 34.5% | 76.0% | Free / Self-Hosted Free / Self-Hosted | 10.09.2026 | ||
| China | 671B (37B active) Reinforcement Learning Reasoning MoE | 128k tokens | 90.8% | 49.2% | 97.3% | $0.55 / 1M tokens $2.19 / 1M tokens | 20.01.2025 | ||
| China | 671B (37B active) MLA + Auxiliary-loss-free MoE | 128k tokens | 88.5% | 42.0% | 89.3% | $0.14 / 1M tokens $0.28 / 1M tokens | 26.12.2024 | ||
| France | 12B Dense Multimodal Vision-Language Transformer | 128k tokens | 69.2% | 28.5% | 62.0% | $0.15 / 1M tokens $0.15 / 1M tokens | 17.09.2024 | ||
| China | 72.7B Dense Dense Transformer Open Weights | 128k tokens | 86.1% | 41.5% | 83.1% | $0.35 / 1M tokens $0.70 / 1M tokens | 19.09.2024 | ||
| China | 32.5B Dense Specialized Code Synthesis Open Weights | 128k tokens | 80.5% | 47.6% | 85.4% | $0.20 / 1M tokens $0.40 / 1M tokens | 12.11.2024 | ||
| United States | 70.6B Dense Dense Transformer Open Weights | 128k tokens | 88.6% | 46.8% | 88.4% | Free / Self-Hosted Free / Self-Hosted | 06.12.2024 | ||
| United States | 405B Dense Ultra-Scale Dense Open Weights | 128k tokens | 88.6% | 38.0% | 84.0% | Free / Self-Hosted Free / Self-Hosted | 23.07.2024 | ||
| United States | 14B Dense Synthetic Data Small Language Model (SLM) | 16k tokens | 84.8% | 39.4% | 80.4% | Free / Self-Hosted Free / Self-Hosted | 12.12.2024 | ||
| United Arab Emirates | 10B Dense Lightweight Dense Open Weights | 32k tokens | 78.5% | 31.2% | 72.0% | Free / Self-Hosted Free / Self-Hosted | 18.12.2024 | ||
| Israel | 398B (94B active) SSM-Transformer (Mamba + Attention) MoE | 256k tokens | 81.2% | 36.5% | 74.8% | $2.00 / 1M tokens $8.00 / 1M tokens | 22.08.2024 | ||
| China | 389B (52B active) Large-Scale MoE Transformer | 256k tokens | 85.2% | 38.5% | 82.4% | Free / Self-Hosted Free / Self-Hosted | 05.11.2024 | ||
| Global | 1.71B Dense On-Device Compact Transformer | 8,192 tokens | 65.2% | 18.4% | 54.0% | Free / Self-Hosted Free / Self-Hosted | 01.11.2024 |
