Skip to content

Products

GPU CloudNVIDIA H200 by the hour.
Dedicated ClustersPreviewPrivate, interconnected GPU clusters for sustained workloads.
SandboxesPreviewIsolated, short-lived environments for developers and AI agents.
Serverless GPUPreviewGPU containers that scale with requests, down to zero.
Serverless InferenceOpen models behind one OpenAI-compatible endpoint.
Dedicated EndpointsReserved serving capacity for one model, billed per GPU-hour.
PlaygroundSend a prompt to any model your workspace can reach.
Model DeploymentsPreviewImmutable deployments with instant promote and rollback.
Fine-tuningLoRA fine-tunes on the open models in our catalog.
Model TrainingPreviewFoundation-model training runs, managed end to end.
EvaluationsPreviewBenchmark suites and model comparisons you can rerun.
DatasetsManaged and synthetic training data, versioned and ready to use.
Data ManagementPreviewDatasets, connections and storage, with lineage across the platform.
PipelinesPreviewVersioned DAGs that automate data, training and deployment.
Data RoomsPreviewSecure document sharing with guests, NDAs and an audit trail.
NokuvaPreviewThe AI-native design editor.
TavocPreviewA Rust-native code platform: IDE and desktop apps.
FoltracComing soonMulti-cloud deployment with autonomous operations.
W0rkTreeComing soonVersion control for agent fleets.
Identity & AccessMembers, roles, policies, SSO and audit for every workspace.
Billing & CostUsage, invoices, credits and budgets in one place.
Logs & MonitoringRequest logs, latency and error rates across services.
Project CanvasPreviewSee and change a project's resources on one live canvas.

Pages

Home/
Platform/platform
All products/products
Pricing/pricing
GPU pricing/pricing#compute
Inference services/services/inference
Model training services/services/model-training
Protocol design/services/protocol-design
Research partnerships/services/research-partnerships
Protocols/protocols
Datasets/datasets
Documentationdocs.legionedge.ai
Changelog/changelog
News/news
Research/research
Blog/blog
Articles/articles
Events/events
Startup Program/programs/startup
Ambassador Program/programs/ambassador
Partner Program/programs/partner
Event Program/programs/events
Company/company
Careers/careers
Partners/partners
Contact/contact
Statusstatus.legionedge.ai
Terms of Service/legal/terms
Privacy Policy/legal/privacy

Changelog

What shipped, release by release.

Every release across Nokuva, Tavoc, Foltrac, and the platform, newest first. Research releases live in the research registry.

releases
12
latest release · 2026
Oct 14
Latest releasePlatform

Organizations, request-based signup and cloud.legionedge.ai

LegionEdge Cloud now runs at cloud.legionedge.ai. New organizations are requested, reviewed by LegionEdge and created on approval.

Platform
  • New

    cloud.legionedge.ai is the console's home, with sign-in at cloud.legionedge.ai/iam/login. app.legionedge.ai and iam.legionedge.ai redirect there with the path kept.

  • New

    Request an organization in three steps: you, your organization, and the plan and products you want.

  • New

    LegionEdge reviews every request. Approval creates the organization and emails its owner an invitation.

  • Improved

    Invitations, password resets and sign-in codes are sent through Mailjet in the LegionEdge design.

2026

  1. Platform
    Platform

    An IAM dashboard for every organization

    Each organization's owner and the administrators they appoint run identity and access for it: members, roles, policies, SSO, SCIM, service accounts, keys and the audit log.

    • New

      Guests, for people outside the organization.

    • New

      My approvals: one inbox for access requests, guest sponsorship, pipeline approval gates and every other decision you've been asked to make.

    • New

      View as opens the console as a member sees it, in its own tab, with a stated reason and a time limit.

    • Improved

      Role templates and a rebuilt role builder for delegating administration.

  2. Platform
    Platform

    Data Rooms and Data Management in preview

    Share documents with guests in Data Rooms, and keep datasets, connections and storage together in Data Management.

    • New

      Data Rooms: folders, resumable uploads, in-browser preview, people and groups, a permission matrix, Q&A and an activity log.

    • New

      Room settings for NDA acceptance, download watermarks, share links, an IP allow-list and legal holds. Guests see only the rooms shared with them.

    • New

      Versioned datasets with preview, quality, lineage and access pages.

    • New

      Connections to Amazon S3, Google Cloud Storage, Azure Blob Storage, Cloudflare R2, Postgres, BigQuery, Snowflake, SFTP, HTTP and Hugging Face, with syncs into datasets.

    • New

      One Storage page for buckets and volumes.

  3. Platform
    Platform

    Pipelines and the Project Canvas in preview

    Automate data work as versioned DAGs, and see and change a project's resources on one live canvas.

    • New

      A visual pipeline editor with validation, dry runs and versioned publishing.

    • New

      Steps for datasets and connections, SQL transforms, schema and expectation checks, PII scans, branches and loops, notifications and approval gates.

    • New

      Schedule, platform-event and webhook triggers, with step-by-step run detail.

    • New

      Project Canvas: staged changes you review and apply together, environments that fork and diff, variables with references, and presence.

  4. Platform
    Platform

    A simpler GPU catalog

    The AMD MI325X and the US Central (Chicago) region leave the catalog. NVIDIA H200 in US South (Dallas) is the default everywhere.

    • Removed

      AMD Instinct MI325X, with its reserved terms and the ROCm image.

    • Removed

      The US Central (Chicago) region, where MI325X ran.

    • Improved

      NVIDIA H200 in US South (Dallas) is the default GPU and region everywhere. It bills on demand at $4.79 per GPU-hour; committed capacity is arranged with sales from the Reservations page.

  5. Nokuva 1.4
    Nokuva

    Streaming layout patches on the canvas

    The canvas now renders Virtual DOM Protocol v2 streaming patches, so designs appear incrementally as the model generates instead of waiting for a complete document.

    • New

      Streaming layout patches — watch structure land on the canvas mid-generation.

    • New

      Accessibility annotation layer from VDOM v2 surfaces in the inspect panel.

    • Improved

      Dirty-subtree layout invalidation: 5,000-node documents stay at 60fps while patching.

    • Fixed

      Token edits no longer drop pending canvas selections during theme switches.

  6. Foltrac 0.9
    Foltrac

    Unified pipelines reach general availability

    One pipeline definition now targets AWS and GCP in production, with per-cloud overrides only where you need them.

    • New

      Single pipeline definition deploys to AWS and GCP with automatic per-provider translation.

    • New

      Dependency-ordered multi-region rollouts, powered by our agent coordination research.

    • Improved

      Rollback plans are pre-validated at deploy time — recovery replays a checked plan.

  7. Tavoc 1.2
    Tavoc

    Scoped context views from W0rkTree

    The IDE now assembles model context through W0rkTree Protocol scoped views, cutting context assembly to under 2% of repository reads on typical tasks.

    • New

      W0rkTree scoped context views — assistance reads what the task needs, not the whole repo.

    • Improved

      Copy-on-write workspaces spin up in under a millisecond.

    • Improved

      Rust analyzer memory usage down 30% on large workspaces.

    • Fixed

      Desktop app no longer re-indexes unchanged dependencies after branch switches.

  8. Nokuva 1.3
    Nokuva

    Multi-agent generation

    Generation is now split across specialized agents — themes, layout, specs, and code each get focused context, which measurably improves output consistency.

    • New

      Specialized agents for themes, layouts, specs, and code conversion.

    • Improved

      Design token suggestions respect the active theme's full token hierarchy.

    • Fixed

      Collaborative undo/redo now resolves correctly across concurrent per-node edits.

  9. Tavoc 1.1
    Tavoc

    First stable update after 1.0

    The first post-1.0 release focuses on stability and startup time, with beta workspaces upgrading in place.

    • Improved

      Cold start time down 40% on Windows and macOS desktop apps.

    • Fixed

      Context assistance no longer stalls on repositories with deep submodule trees.

    • Fixed

      Template workspaces restore correctly after an interrupted first sync.

  10. Foltrac 0.8
    Foltrac

    Autonomous rollback policy engine

    Foltrac now watches error rates, latency distributions, and resource envelopes against the previous release's baselines, and rolls back on sustained, attributable regressions.

    • New

      Baseline-relative regression detection — no hand-tuned absolute thresholds.

    • New

      Asymmetric rollback policy: moderate evidence reverts, strong evidence is needed to keep a suspect release.

    • Improved

      Deploy logs record the pre-validated inverse plan alongside every release.

  11. Platform
    Platform

    W0rkTree goes open source

    The Rust version control core behind our tooling is now public under GPL 3.0, with the W0rkTree Protocol spec published alongside it.

    • New

      W0rkTree core repository public on GitHub under GPL 3.0.

    • New

      W0rkTree Protocol spec v1 in the research registry.

    • Improved

      Contribution guide and conformance test suite for third-party implementations.

Keep reading

Want the story behind the releases?

Launches and announcements are in News; the engineering behind them is on the Blog.