Best Open-Weight LLMs
Open-weight LLMs publish their model weights, so you can run them on your own infrastructure, fine-tune them, and avoid vendor lock-in. The table below lists open-weight models ranked by context window and price. These models give you the most control over deployment, privacy, and cost.
Ranked list (top 25)
| # | Model | Input / 1M | Output / 1M | Context |
|---|---|---|---|---|
| 1 | Llama 4 Scout 17B Instruct (US) Amazon BedrockOpen weights | $0.17 | $0.66 | 10M |
| 2 | DeepSeek V4.1 Flash Marathon SubconsciousOpen weights | $0.14 | $0.28 | 5M |
| 3 | GLM-5.3 Marathon SubconsciousOpen weights | $1.40 | $4.40 | 5M |
| 4 | Meta: Llama 4 Scout MetaOpen weights | $0.10 | $0.30 | 1.3M |
| 5 | DeepSeek V4.1 Flash 1017 (sciforium) LLM GatewayOpen weights | $0.15 | $0.60 | 1.1M |
| 6 | MiMo M2.5 Vercel AI GatewayOpen weights | $0.12 | $0.24 | 1.1M |
| 7 | DeepSeek V4 Flash (NovitaAI) LLM GatewayOpen weights | $0.14 | $0.28 | 1.1M |
| 8 | Xiaomi: MiMo-V2.5 XiaomiOpen weights | $0.14 | $0.28 | 1.1M |
| 9 | DeepSeek V4.1 Flash (DeepSeek) LLM GatewayOpen weights | $0.15 | $0.60 | 1.1M |
| 10 | DeepSeek V4.1 Flash 1017 (DeepSeek) LLM GatewayOpen weights | $0.15 | $0.60 | 1.1M |
| 11 | Xiaomi: MiMo-V2.5-Pro XiaomiOpen weights | $0.44 | $0.87 | 1.1M |
| 12 | deepseek-ai/DeepSeek-V4-Pro SiliconFlow (China)Open weights | $1.79 | $3.59 | 1M |
| 13 | Beam 1.0 Vercel AI GatewayOpen weights | $0.00 | $0.00 | 1M |
| 14 | Beam Free OpenCodeOpen weights | $0.00 | $0.00 | 1M |
| 15 | Kimi K3 (Free) ZenMuxOpen weights | $0.00 | $0.00 | 1M |
| 16 | MiMo-V2.5 (Free) KenariOpen weights | $0.00 | $0.00 | 1M |
| 17 | Nemotron 3.5 Super VL 120B A12B NvidiaOpen weights | $0.00 | $0.00 | 1M |
| 18 | Xiaomi MiMo-V2.5 (free) AIHubMixOpen weights | $0.00 | $0.00 | 1M |
| 19 | Xiaomi MiMo-V2.5-Pro (free) AIHubMixOpen weights | $0.00 | $0.00 | 1M |
| 20 | DeepSeek V4 Flash (Runware) LLM GatewayOpen weights | $0.08 | $0.15 | 1M |
| 21 | Coding Xiaomi MiMo-V2.5 AIHubMixOpen weights | $0.08 | $0.16 | 1M |
| 22 | DeepSeek V4 Flash 0731 (Thinking) NanoGPTOpen weights | $0.05 | $0.16 | 1M |
| 23 | DeepSeek V4 Flash 0731 (Deep Infra) Eden AIOpen weights | $0.06 | $0.18 | 1M |
| 24 | DeepSeek: DeepSeek V4 Flash 0423 DeepSeekOpen weights | $0.09 | $0.18 | 1M |
| 25 | Laguna S 2.1 Thinking NanoGPTOpen weights | $0.10 | $0.20 | 1M |
Prices are per 1M tokens (USD); confirm with the provider. Updated regularly.
More LLM rankings
Frequently asked questions
What is an open-weight LLM?
An open-weight LLM is a model whose trained weights are publicly available for download, letting you self-host, fine-tune, and inspect it. This differs from fully open-source (which also releases training code and data) and from closed API-only models.
What is the best open-weight LLM?
Among open-weight models in our dataset, Llama 4 Scout 17B Instruct (US) ranks highly on context window and pricing. The best choice depends on your hardware budget and task.
Can I run open-weight LLMs locally?
Yes — open-weight models can run on your own GPUs or via inference providers. Smaller models run on a single GPU; larger ones need multi-GPU setups or quantization.