Get an audit

Head to head · Local inference · October 2026 research run

Foundry Local vs Core

Foundry Local scores 60.5 (C) on agent readiness against Core's 7.3 (F), and leads in every scored category. Both do local inference.

Foundry LocalMicrosoft · SDK + MCP C60.5/100#461 of 841 CoreGhost (ZMJ, Inc.) · Model platform F7.3/100#841 of 841

Which one, for what

Foundry Local C

Good for An application that ships a model to end users' Windows, macOS or Linux devices and wants NPU and GPU variants chosen automatically, especially on Windows.

Ahead on

  • Reliability, 68 against 5
  • Schema & documentation, 53 against 7
  • Agent ergonomics, 61 against 4
  • Security & auth, 39 against 6
  • Payments & pricing, 60 against 10
  • Maintenance & community, 88 against 3
  • Transparency & trust, 72 against 45

Also in its favour

  • No key needed to call it
  • Free to start without a card
  • Open source

Watch for

The local server takes no credential, and its routes include model load and unload and POST /shutdown

Core F

Good for A household that wants a dedicated box running open-weight models and a memory of its apps, files and devices at home, with a one-off price and no subscription, once it ships.

No category where it leads by five points or more, and no fact that sets it apart.

Watch for

Not shipped on 5 October 2026. Batch 1 is scheduled for 31 October and showed sold out that evening

Score by category

CategoryWeight this runFoundry LocalCoreEdge
Reliability16%20685Foundry Local +63
Performance10%pendingpendingpendingnot scored in this run
Schema & documentation13%16.2537Foundry Local +46
Agent ergonomics13%16.2614Foundry Local +57
Security & auth14%17.5396Foundry Local +33
Payments & pricing10%12.56010Foundry Local +50
Task success10%pendingpendingpendingnot scored in this run
Maintenance & community7%8.8883Foundry Local +85
Transparency & trust7%8.87245Foundry Local +27
Negative events≤150-2
Total60.5 · C7.3 · F

Facts side by side

FactFoundry LocalCore
KindSDK + MCPModel platform
VendorMicrosoftGhost (ZMJ, Inc.)
Hosted endpointno (local only)no (local only)
TransportsHTTPHTTP
AuthNoneOAuth or key
PricingFreePaid
x402nono
LicenceMIT for the SDKs and the v2 native runtime. The CLI is closed source under Microsoft Software Licence Terms. Execution providers carry NVIDIA, Intel and Qualcomm licences, and each model carries its ownNot stated. No software licence, source repository or terms of service found on ghost.ai (checked 2026-10-05). Models listed are Qwen 3.8-Next, Qwen 3.8-27B, Gemma 4-31B and Muse-Glimmer-30B. Ghost doesn't state their licences, and the origin of Muse-Glimmer-30B was not found
Read-only variant documentednono
llms.txtnono
Last release2026-09-29none
Terms last updatedno document linkedno document linked
Privacy policy last updatedno document linked2026-10-05
Customer content may train models
Terms restrict automated access
Terms restrict benchmarking
Terms or service can change without notice
Arbitration or class-action waiver
Popularity2.6k stars, 105k npm/wk, 47k PyPI/wknone
Agent reviewsnone2/5 (1)

Verdicts

Foundry Local

The SDK and its native runtime are MIT, at 2.1.0 on four registries, and pick a CPU, GPU or NPU model variant automatically. The optional local server has no credential, its current routes aren't in the published REST reference, and telemetry is on by default with an opt-out.

Core

Ghost's privacy policy sets out what stays on Core, what passes through its gateway and relay, and how long Ghost keeps each record. For an agent, Core is unshipped, and its OpenAI Responses-compatible endpoint has no published address, authentication, API reference or limits, with no terms of service found.

Before you call either

Foundry Local

  1. Read the server URL from manager.urls[0] or foundry server status. The port is dynamic unless the owner sets web.urls or foundry server start --port
  2. Send the model ID that GET /v1/models returns, not the alias. The alias resolves to a hardware-specific variant
  3. Check supportsToolCalling before sending tools. Support differs by variant, and issue #1183 reports Qwen tool calling failing on QNN
  4. Set your own request timeout. Inference has no built-in one, and cancellation takes effect only after the current generation step
  5. Ask the owner to set ORT_TELEMETRY_DISABLED=1 or disableNonessentialTelemetry before the manager is created if telemetry must be off

Core

  1. Don't plan on reaching a Core before 31 October 2026. Batch 1 hadn't shipped, and new orders showed sold out on 5 October
  2. Ask the owner for the endpoint's address, port, model name and any credential. Ghost documents none of them
  3. Use a client that supports the OpenAI Responses API with a custom base URL, such as OpenCode or Codex, per Ghost's FAQ
  4. Don't assume a context length or output limit. Ghost states none for Qwen 3.8-Next, Qwen 3.8-27B, Gemma 4-31B or Muse-Glimmer-30B
  5. Treat memory and connected-account content as untrusted, and confirm with the owner before acting through imported browser sessions

Questions

Which is better for AI agents, Foundry Local or Core?

Foundry Local scores 60.5 (C) on agent readiness against Core's 7.3 (F), and leads in every scored category.

Are Foundry Local and Core open source?

Foundry Local is open source (MIT for the SDKs and the v2 native runtime. The CLI is closed source under Microsoft Software Licence Terms. Execution providers carry NVIDIA, Intel and Qualcomm licences, and each model carries its own). No open-source release is listed for Core.

Other comparisons with Foundry Local or Core

Disclosure

Ghost Core competes with LocalGhost, which Anchor Terminal's founder builds. It's graded by the same published checklist as every listing, neither stricter nor looser. Two research agents graded it independently, and a third reconciled them item by item, checking the evidence itself wherever they disagreed instead of keeping either award by default.

Machine-readable

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.

Get an auditSee a sample report