PlatformRoadmap
Roadmap
Four stages, one direction: from secure inference to custom training on your own data.
Last updated September 20, 2026
Not stored by defaultEU · USA · LATAMOpenAI-compatible
V1Available now
Secure & private inference
- Region-pinned inference on dedicated capacity — EU, USA, LATAM
- Prompts and outputs are not stored by default — no sharing, no training
- Qwen 3.8 (1M context) under the 99.999% SLA
- Clef decisions and embedding models under the 99.999% SLA
- Whisper
Explore this stage
V2Coming soon
Connection & intelligence
- Knowledge bases
- Connectors
- Knowledge Intelligence
- Advanced Guardrails
- MCP & tool access
- Agent runtime
- Multi-agent orchestration
- Brain: persistent organizational knowledge — 10–50 GB, encrypted, region-pinned, isolated to the organization, available across plans
- Memory: persistent context for users, agents and tasks — available across plans
Explore this stage
V3Planned
Your models, your infrastructure
- Bring your own models (BYOM)
- Dedicated, customer-specific models
- Dedicated, customer-specific infrastructure
Explore this stage
V4Planned
Custom training
Custom training on your organization's data
Explore this stage
How to read this roadmap
Stages follow the approved product direction. V1 is live in each region as capacity lands. V2 and V3 are coming soon and V4 is planned: nothing beyond V1 is live today, and no stage carries a date commitment — Qwen 3.8 (1M context), GLM 5.3, GLM 5.3 Flash and DeepSeek V4.1 Flash 1017 run today under the 99.999% SLA.
Explore the platform
Start with what is live
V1 is available today in your region.