Model Rankings
Every live model on Frontière AI is ranked by two criteria: Best Value (capability-to-price ratio — bigger, smarter models at lower prices rank higher) and Best Performance (raw capability, verified score, multimodality, and context length). Sovereign EU models get a +30% boost in the value ranking.
Rankings
26B · 6/6 ✓ · EU sovereign · €0.0168/100k in
27B · 4/6 ✓ · Fast access (US) · €0.0135/100k in
119B · 5/6 ✓ · EU sovereign · €0.0278/100k in
78B · 6/6 ✓ · EU sovereign · €0.0252/100k in
117B · 6/6 ✓ · EU sovereign · €0.0252/100k in
763B · 6/6 ✓ · EU sovereign · €0.1008/100k in
1.6T · 5/6 ✓ · Fast access (US) · €0.2453/100k in
2.8T · 3/6 ✓ · Fast access (US) · €0.4536/100k in
1T · 4/6 ✓ · EU sovereign · €0.2514/100k in
753B · 6/6 ✓ · EU sovereign · €0.3024/100k in
| # | Model | Size | Price (input) | Jurisdiction | Verified Score |
|---|---|---|---|---|---|
| 1 | Gemma 4 26B (A4B MoE) | 26B | €0.0168 | EU sovereign | 6/6 |
| 2 | Gemma 3 27B (Google) | 27B | €0.0135 | Fast access (US) | 4/6 |
| 3 | Mistral Small 4 119B | 119B | €0.0278 | EU sovereign | 5/6 |
| 4 | Kolibri 1 (Aleph Alpha) | 78B | €0.0252 | EU sovereign | 6/6 |
| 5 | GPT-OSS 120B (OpenAI) | 117B | €0.0252 | EU sovereign | 6/6 |
| 6 | Qwen3.8 27B (multimodal) | 27.8B | €0.0252 | EU sovereign | 6/6 |
| 7 | Mistral Small 3.2 24B | 24B | €0.0252 | EU sovereign | 5/6 |
| 8 | Qwen3 Coder 30B | 30.5B | €0.0336 | EU sovereign | 6/6 |
| 9 | Qwen3.6 35B A3B (multimodal) | 36B | €0.042 | EU sovereign | 6/6 |
| 10 | Pixtral 12B (vision) | 12B | €0.0336 | EU sovereign | 5/6 |
| 11 | DeepSeek V4 Flash 0731 | 284B | €0.0672 | EU sovereign | 6/6 |
| 12 | DeepSeek V4.1 Flash (552B MoE, 1M context) | 763B | €0.1008 | EU sovereign | 6/6 |
| 13 | Qwen3.5 397B (multimodal) | 397B | €0.1008 | EU sovereign | 6/6 |
| 14 | GLM-5.3 Flash (Zhipu/Z.ai) | 320B | €0.0689 | Fast access (US) | 0/6 |
| 15 | Qwen3 235B (instruct) | 235B | €0.126 | EU sovereign | 5/6 |
| 16 | Apertus v1.5 70B (Swiss open model) | 72B | €0.1294 | EU sovereign | 5/6 |
| 17 | Mistral Large 4 (preview, 512K context) | 1T | €0.2514 | EU sovereign | 4/6 |
| 18 | Llama 3.3 70B | 70B | €0.1512 | EU sovereign | 5/6 |
| 19 | DeepSeek V4 Pro (1.6T MoE) | 1.6T | €0.2453 | Fast access (US) | 5/6 |
| 20 | GLM-5.2 (Zhipu/Z.ai) | 753B | €0.3024 | EU sovereign | 6/6 |
| 21 | GLM-5.3 (Zhipu/Z.ai) | 753B | €0.3024 | EU sovereign | 6/6 |
| 22 | Kimi K3 (2.8T parameters, 1M context) | 2.8T | €0.4536 | Fast access (US) | 3/6 |
| 23 | Muse Spark 1.1 (1M context) | ? | €0.205 | Fast access (US) | 3/6 |
| 24 | Muse Spark 1.2 (1M context) | ? | €0.205 | Fast access (US) | 3/6 |
| 25 | Qwen3.8 Flash (Qwen) | ? | €0.0252 | Fast access (US) | 3/6 |
Two rankings: Best Value computes a capability-to-price score (log-scale parameters × sovereignty +30% × context bonus × multimodality × verified score) divided by input price. Best Performance ranks by raw capability (log-scale parameters, with a +5% bonus per confirmed Verified criterion, multimodality and context length). The goal: help you pick the right model for your use case, not just the cheapest one.