AkashML
Decentralised, per-token inference for open models such as Qwen, GLM, Kimi and Llama, running on the Akash Network.
Overview
AkashML models are available through Nautili. Requests to them use the same routing, governance and observability as every other request.
Models
- Gemma 4 26B A4B google/gemma-4-26b-a4b-itView model: Gemma 4 26B A4B
- Meta: Llama 3.3 70B Instructmeta-llama/llama-3.3-70b-instructView model: Meta: Llama 3.3 70B Instruct
- Kimi K3moonshotai/kimi-k3View model: Kimi K3
- gpt-oss-120bopenai/gpt-oss-120bView model: gpt-oss-120b
- gpt-oss-20bopenai/gpt-oss-20bView model: gpt-oss-20b
- Qwen3.6 35B A3Bqwen/qwen3.6-35b-a3bView model: Qwen3.6 35B A3B
- Qwen3.8 27Bqwen/qwen3.8-27bView model: Qwen3.8 27B
- GLM 5.3z-ai/glm-5.3View model: GLM 5.3
Integration
Point your existing SDK or tool at your Nautili endpoint and key. Nautili routes requests to AkashML models within your policies.
Managed: use Nautili-issued credentials.
Connect your stack to Nautili
Route across supported providers through one control layer.