Pyyan / Open-Weight Models
Open-Weight Models
38 entriesDownloadable weights you can self-host. · ranked by benchmark performance and adoption · verified 29 Sept 2026
These also appear on the full language models page →
| # | Name | Context | Input ($/Mtok) | Output ($/Mtok) | Licence | Max output | Status |
|---|---|---|---|---|---|---|---|
| 1 | Qwen3.8-Flash-Next Alibaba | 262K native, ~1M extended | $0.15 | $0.47 | Open weights | 32K | current |
| 2 | DeepSeek V4.1 Flash DeepSeek | 1M | $0.3 | $1.2 | MIT | 256K | current |
| 3 | DeepSeek V4-Pro DeepSeek | — | — | — | MIT | — | current |
| 4 | XMiMo-V2.6-Pro Xiaomi | 1M | Weights only | Weights only | MIT | Not published | current |
| 5 | XMiMo-V2.6-Flash Xiaomi | 1M | Weights only | Weights only | MIT | Not published | current |
| 6 | SAtria Dawn Preview Shanghai AI Laboratory | 256K | Not published | Not published | MIT | Not published | current |
| 7 | DeepSeek V4 Flash Vision Exp DeepSeek | 128K | $0.44 | $1.32 | Open weights | 32K | deprecated |
| 8 | TInkling Thinking Machines Lab | 1M | Weights only | Weights only | Apache 2.0 | Not published | current |
| 9 | Kimi K3 Moonshot AI | — | — | — | Open weights | — | current |
| 10 |  GLM-5.3-FlashZhipu AI | 1M | $0.15 | $0.5 | MIT | 32K | current |
| 11 | GLM-5.3 Zhipu AI | 1M | $1.4 | $4.4 | Open weights | 32K | current |
| 12 | GLM-5.2 Zhipu AI | — | — | — | Open weights | — | current |
| 13 | Qwen3-235B-A22B Alibaba | — | — | — | Apache-2.0 | — | current |
| 14 | Llama 5 Meta | — | — | — | Llama licence | — | current |
| 15 |  Hy4 previewTencent | 1M | $0.834 | $2.501 | Apache 2.0 | 64K | current |
| 16 | Hunyuan 3.0 Tencent | 256K | — | — | Apache-2.0 | — | current |
| 17 | Qwen3.8-27B Alibaba | 262K native, ~1M extended | Weights only | Weights only | Apache 2.0 | 32K | current |
| 18 | Muse Glimmer Meta | 131K | Weights only | Weights only | Apache 2.0 | 16K | current |
| 19 | Qwen3-Coder-480B Alibaba | — | — | — | Apache-2.0 | — | current |
| 20 | MiniMax M3 MiniMax | — | — | — | Open weights | — | current |
| 21 |  gpt-oss-120bOpenAI | — | — | — | Open weights | — | current |
| 22 | Nemotron 3.5 Lightning NVIDIA | 128K | Weights only | Weights only | OpenMDW-1.1 | 32K | current |
| 23 | Nemotron 3 Ultra NVIDIA | 1M | — | — | NVIDIA Open Model | — | current |
| 24 | Kimi K2.6 Moonshot AI | — | — | — | Open weights | — | current |
| 25 | GLM-5.1 Zhipu AI | — | — | — | Open weights | — | current |
| 26 | DeepSeek V3.2 DeepSeek | — | — | — | MIT | — | current |
| 27 | Llama 4 Maverick Meta | — | — | — | Llama licence | — | current |
| 28 | Mistral Large 3 Mistral | — | — | — | Apache-2.0 | — | current |
| 29 |  Granite 4.2IBM | 131K, 512K extension claimed | Weights only | Weights only | Apache 2.0 | 32K | current |
| 30 |  Phi-4Microsoft | — | — | — | MIT | — | current |
| 31 | Gemma 3 27B | — | — | — | Gemma licence | — | current |
| 32 | TFalcon TII | — | — | — | Open weights | — | current |
| 33 | ERNIE 4.5 Baidu | — | — | — | Open weights | — | current |
| 34 | gpt-oss-20b OpenAI | — | — | — | Open weights | — | current |
| 35 | Phi-4-mini Microsoft | — | — | — | MIT | — | current |
| 36 | Gemma 3 4B | — | — | — | Gemma licence | — | current |
| 37 |  Step 3.5 FlashStepFun | — | — | — | Open weights | — | current |
| 38 | SSarvam Sarvam AI | — | — | — | Open weights | — | current |
#1
Qwen3.8-Flash-Next
Alibaba
125B
total, 6B active
- 125B total parameters with only 6B active per token
- Alibaba describes it as a preview of the Qwen4 architecture
- Top of the Hugging Face trending list, with a GGUF build close behind it
Best foropen weights at speed
#2
DeepSeek V4.1 Flash
DeepSeek
90.6%
Terminal-Bench 2.1
- 552B total, 8B active while reading the prompt and 16B while generating, with a 1M token context
- DeepSeek says it beats V4-Pro; V4-Pro requests route to it from 14 September 2026
- MIT licence, native image input, and a KV cache about a quarter the size of V4-Flash's
Best forcheap agentic coding with images