AI 모델 인텔리전스

기능 · 2026-11-12

오픈 웨이트 AI 모델

라이선스 하에 학습된 가중치를 공개하는 AI 모델 비교 — 셀프 호스팅, 파인튜닝, 규제 환경용.

이게 뭔가요?

  • 오픈 웨이트 모델은 가중치를 공개합니다(주로 Apache 2.0, MIT 또는 오픈 연구 라이선스).
  • 참고: 오픈 웨이트 ≠ 오픈 소스 — 학습 데이터와 코드는 보통 공개되지 않습니다.

왜 중요한가

  • 자체 GPU에서 셀프 호스팅, 비공개 데이터로 파인튜닝, 오프라인 운영, 규제 환경 배포가 가능합니다.
  • 표에는 가장 저렴한 API 호스트 가격이 표시됩니다 — 자체 하드웨어가 있다면 API 비용 없이 동일 모델을 운영할 수 있습니다.

이 기능을 지원하는 모델 931개

모델벤더입력 / 1M출력 / 1M컨텍스트제공자
Qwen3.8 27BAlibaba (Qwen)$0.001$0.0011.05M51
Ambient Smallambient$0.001$0.001101K1
Whisper Large v3OpenAI$0.002$0.0024483
Voxtral Mini 3Bprivatemode-ai$0.005Unknown32K1
Multi-QA-mpnet-base-dot-v1digitalocean$0.009Unknown5121
All-MiniLM-L6-v2digitalocean$0.009Unknown2561
Qwen3 Embedding 8BAlibaba (Qwen)$0.010Unknown33K8
Qwen3 Embedding 4BAlibaba (Qwen)$0.010Unknown33K4
BGE Reranker V2 M3novita-ai$0.010Unknown8K1
BGE M3novita-ai$0.010Unknown8K1
BGE Reranker v2 M3digitalocean$0.010Unknown8K1
Nomic Embed Text v1.5tinfoil$0.016Unknown8K1
Whisper large-v3privatemode-ai$0.016Unknown4481
Llama 3.2 1B InstructMeta$0.010$0.01060K6
Qwen3 Embedding 0.6BAlibaba (Qwen)$0.010$0.01033K3
Prompt Guard 2 86MMeta$0.010$0.0105123
Llama Prompt Guard 2 22MMeta$0.010$0.0105123
E5 Large v2digitalocean$0.020Unknown5121
BGE M3digitalocean$0.020Unknown8K1
GLiNER 2.5 Basepioneer$0.030Unknown4K1
GLiNER2 Multi Largepioneer$0.030Unknown8K1
GLiNER-2.5-Decidepioneer$0.030Unknown8K1
GLiNER2 Basepioneer$0.030Unknown2K1
GLiNER2 Privacy Filter PII (Multi)pioneer$0.030Unknown8K1
GLiNER2 Largepioneer$0.030Unknown8K1
GLiNER2 Multipioneer$0.030Unknown8K1
GLiNER 2.5 Multipioneer$0.030Unknown4K1
Llama 3.2 3B InstructMeta$0.020$0.020131K7
PaddleOCR VLnovita-ai$0.020$0.02016K1
Gemma 4 26B A4B ITGoogle$0.010$0.031262K23
Llama-3.1-8B-InstructMeta$0.020$0.030131K14
Meta Llama 3.1 8B Instruct TurboMeta$0.020$0.030128K1
Mistral NemoMistral$0.029$0.030131K6
DeepSeek OCR 2DeepSeek$0.030$0.0308K3
Mistral NemoMistral$0.020$0.04016K3
Ling 3.0 Tinyopenrouter$0.015$0.045262K1
Ministral 3Bazure-cognitive-services$0.040$0.040128K1
Ministral 3Bazure$0.040$0.040128K1
Ling 3.0 Flashvercel$0.021$0.063256K1
Ling 3.0 Flashopenrouter$0.021$0.063262K1
Gemma 3 4B ITGoogle$0.040$0.050131K11
Llama 3 8B LunarisMeta$0.040$0.0508K3
GTE Large (v1.5)digitalocean$0.090Unknown8K1
Qwen3.5 9BAlibaba (Qwen)$0.040$0.060262K21
Qwen2.5 7B Instruct (Deep Infra)Alibaba (Qwen)$0.050$0.05033K3
gpt-oss-20bOpenAI$0.018$0.090131K32
Llama-3.2-11B-Vision-Instruct (Deep Infra)Meta$0.055$0.055131K4
Llama-Guard-3-8B (Cloudflare)Meta$0.055$0.055131K3
Qwen3.5 4BAlibaba (Qwen)$0.040$0.070262K5
Nex N2.5 Mininano-gpt$0.025$0.100262K1
Nex-N2.5-Miniopenrouter$0.025$0.100262K1
Granite 4.0 H Microcloudflare-workers-ai$0.017$0.112131K1
Mistral Small 3Mistral$0.050$0.08033K6
Llama 3.1 8BMeta$0.050$0.080131K2
DeepSeek V4 Flash 0731DeepSeek$0.045$0.0901.05M54
Baichuan M2 32Bnovita-ai$0.070$0.070131K1
Nemotron 3.5 Lightning 30B A3BNVIDIA$0.025$0.120262K5
DeepSeek-V4.1-FlashDeepSeek$0.030$0.1201M72
DeepSeek V4.1 Flash 1017DeepSeek$0.030$0.1201.05M60
LFM2-24B-A2Btogetherai$0.030$0.12033K1

전체 931개 중 상위 60개 표시. 추가 필터링은 전체 목록을 이용하세요.

Frequently asked questions

How many AI models support 오픈 웨이트?

931 canonical models in our database currently support 오픈 웨이트. The list is regenerated on every data refresh, so it always reflects the latest releases tracked in our catalogue.

What is the cheapest model with 오픈 웨이트?

Qwen3.8 27B from Alibaba (Qwen) is currently the lowest-priced option, at $0.001 per 1M input tokens and $0.001 per 1M output tokens. The full table above is sorted price-ascending.

Which model with 오픈 웨이트 has the largest context window?

Llama 4 Scout 17B Instruct (US) (Meta) leads on context at 10M tokens. This may matter if you also need long-document understanding alongside 오픈 웨이트.

Which models are available on the most providers?

Production-readiness usually correlates with how many independent providers host the same weights. The top three by provider count are: GLM-5.3-Flash (92), GLM-5.3 (92), GLM-5.2 (90).

How is 오픈 웨이트 different from a regular LLM?

Open-weight models publish their trained weights publicly. You can self-host on your own GPUs, fine-tune on private data or run offline. Note: open weights ≠ open source — training data and code are usually not released.

How often is this list updated?

Daily. Our data pipeline syncs once a day, regenerates the canonical model list, and rebuilds these pages so newly released models appear within 24 hours.

마지막 업데이트:

Prices in USD per 1M tokens. Unknown means the provider does not publish per-token pricing.

Pricing and capabilities are refreshed daily and reconciled against each provider's official documentation. Always verify critical production decisions with the provider directly.