-
Notifications
You must be signed in to change notification settings - Fork 0
Roadmap
This page gives a concise project roadmap.
XAIOS has a bootable QEMU prototype and deterministic OS/runtime fixtures. Its portable model-v2 and inference-engine boundaries are foundational interfaces; real-model tokenizer, logits and decode parity remain incomplete.
The detailed milestone checklist lives in Project Tracker. Current execution status lives in the XAIOS GitHub Project.
Authoritative delivery order:
- Finish the XAIOS platform milestone. It is the only active workstream.
- Implement and validate Qwen 3.6 27B. It is next but blocked on XAIOS.
- Keep Kimi K3, DeepSeek V4 Flash 0731, and GLM 5.2 queued unless explicitly reprioritized. DeepSeek is also blocked on authoritative source identity.
Platform-specific Apple, Intel, Xeon, and ARM work is part of the XAIOS completion gate when required by its correctness and hardware-readiness criteria; it is not a competing model workstream.
| Model | Status | Next gate |
|---|---|---|
| Qwen 3.6 27B | Interface only; next but blocked | Start after XAIOS, then complete tokenizer, logits, deterministic decode and physical-hardware parity. |
| Kimi K3 | Interface/roadmap only; backlog | Queued after Qwen; complete text correctness, then separate multimodal parity. |
| DeepSeek V4 Flash 0731 | Roadmap only | Verify and pin the exact official release and architecture sources. |
| GLM 5.2 | Roadmap only | Pin official sources and define a separate validated adapter. |
See Model Support Roadmap. These labels record planned work and do not mean that model inference is currently available.
| Phase | Focus | Deliverables |
|---|---|---|
| 0 | Foundation | repo layout, build scripts, QEMU scripts, tests |
| 1 | Bootable server core | UEFI boot, serial logging, kernel entry, PMM, VMM |
| 2 | Device and user base | interrupts, timers, user init, VirtIO block/network |
| 3 | AI Cell runtime | core leases, arenas, manifests, telemetry |
| 4 | Low-latency networking | per-core TCP/UDP, queue ownership, latency counters |
| 5 | CPU AI runtime | shared weights, private KV/cache, tokenizer boundary |
| 6 | Intel Desktop | P-core/E-core policy, NVMe, PCIe NIC, benchmarks |
| 7 | Intel Xeon | NUMA, server NICs, many AI Cells, admission control |
| 8 | ARM/NVIDIA | AArch64, GIC, SMMU, cluster policy |
| 9 | Developer preview | sample app agent, benchmark scripts, release notes |
This page defines the GitHub Wiki navigation sidebar.
- Architecture
- AI Cells
- CPU AI Runtime
- App Agents
- Memory System
- Networking
- Scheduler and Core Isolation
- Filesystem and Storage
- Driver Model
- Security Model
- Build System
- Build System
- Project Tracker
- Implementation Plan
- QEMU Full OS Core Workdown
- QEMU 100 Completion Plan
- Example Apps
- Codex Work Packages
- Testing and Benchmarking