Frontière Verified

Independent Model Quality Benchmarks

Frontière AI runs a 6-criteria verification on every model in our catalog. Each model is independently checked for license compliance, provenance, security, GDPR alignment, benchmark quality, and API availability. A model is Frontière Verified when it passes all six checks.

27

Live models

11

Frontière Verified

14

EU sovereign

13

Fast access

Live models — verification status

ModelScoreJurisdictionlpsgba
DeepSeek V4 Flash 0731 (1M context)Frontière VerifiedEU sovereign
DeepSeek V4.1 Flash (552B MoE, 1M context)Frontière VerifiedFast access
GLM-5.2 (Zhipu/Z.ai)Frontière VerifiedEU sovereign
GPT-OSS 120B (OpenAI)Frontière VerifiedEU sovereign
GPT-OSS 20B (OpenAI)Frontière VerifiedEU sovereign
Mistral Small 3.2 24BFrontière VerifiedEU sovereign
Qwen3 32BFrontière VerifiedEU sovereign
Qwen3 Coder 30BFrontière VerifiedEU sovereign
Qwen3.5 397B (multimodal)Frontière VerifiedEU sovereign
Qwen3.5 9B (fast, budget)Frontière VerifiedEU sovereign
Qwen3.6 27B (multimodal)Frontière VerifiedEU sovereign
Kimi K3 (2.8T parameters, 1M context)5/6 criteria passedFast access
Llama 3.3 70B5/6 criteria passedEU sovereign
Qwen2.5-VL 72B (vision)5/6 criteria passedEU sovereign
Qwen3.8 27B (multimodal)5/6 criteria passedEU sovereign
DeepSeek R1 (671B MoE)0/6 criteria passedFast access
DeepSeek R1 (Distill Qwen 14B)0/6 criteria passedFast access
DeepSeek R1 (Distill Qwen 32B)0/6 criteria passedFast access
DeepSeek V4 Pro (345B MoE)0/6 criteria passedFast access
Gemma 3 27B (Google)0/6 criteria passedFast access
GLM-5.3 (Zhipu/Z.ai)0/6 criteria passedFast access
GLM-5.3 Flash (Zhipu/Z.ai)0/6 criteria passedFast access
Muse Spark 1.1 (1M context)0/6 criteria passedFast access
Muse Spark 1.2 (1M context)0/6 criteria passedFast access
Qwen3 235B (instruct)0/6 criteria passedEU sovereign
Qwen3.8 Flash (Qwen)0/6 criteria passedFast access
Qwen3.8 Max (2.4T, 1M context)0/6 criteria passedFast access

Ranking methodology

Models are sorted by: 1) Frontière Verified score (6 criteria: license, provenance, security, GDPR, benchmarks, API), 2) EU sovereign status (sovereign providers ranked above non-sovereign), 3) Price efficiency (lower per-token price ranked higher). Unlike usage-based leaderboards that measure popularity, our rankings measure quality and compliance.

Frequently asked questions

What is Frontière Verified?
Frontière Verified is Frontière AI's independent verification program. Each model in our catalog is evaluated against 6 criteria: license verified, provenance verified, security analyzed, GDPR assessed, benchmarked, and API available. A model earns the Frontière Verified badge when it passes all six checks.
How are models benchmarked?
Benchmarked models run our quality test (25 self-contained QA questions across general knowledge, math, reasoning, and language), latency test (first token time and tokens per second via streaming), and throughput test (sustained tokens/s over a 500+ word generation). A model passes when quality >= 72%, FTT <= 5s, throughput >= 20 tok/s.
What does provenance verified mean?
Provenance verified means we confirmed the model's weights exist on Hugging Face, the publisher is confirmed, and it is not a dubious fork. We verify the exact model ID against the provider's catalog.
What does GDPR assessed mean?
GDPR assessed means we evaluated the hosting provider's data retention policy, sub-processors, and data processing agreements. EU sovereign providers (OVHcloud, Scaleway, self-hosted) are automatically assessed as compliant with EU data residency requirements.
Why does some models have fewer criteria checked?
When a model is newly added to our catalog, we prioritize checking provenance, GDPR, and API availability first. License, security analysis, and benchmarks follow as they take more time to complete. We publish the partial score rather than hiding the model.

Ready to test verified models?

Create an account