Compliance

Terms of Service

Pending legal review — placeholder Effective June 4, 2026GWMM LLC · gwmmai.com
Placeholder text. Specific performance guarantees, liability caps, and indemnification language below are drafts for legal review — they do not constitute final commitments. The architectural guarantees described elsewhere on this site (Zero Data Retention, single-tenant isolation, advertised precision) are operational facts, not legal warranties.

01 Acceptance

These Terms of Service (“Terms”) govern your use of the GWMM direct-route AI inference endpoint (the “Service”). The Service is operated by GWMM LLC (“GWMM”, “we”, “us”) and is reached by you via OpenRouter or via GWMM direct connection. By sending requests to the GWMM endpoint you agree to these Terms and, when using the OpenRouter route, to OpenRouter’s own terms of service.

02 Service Description

GWMM provides access to open-weight language models (currently Gemma 4 12B IT, with Gemma 4 26B MoE in provisioning) on a dedicated, single-tenant inference node. Each request runs on a dedicated instance with page-level KV isolation; no shared queue, no cross-tenant scheduling.

  • OpenAI-compatible chat completions endpoint at /v1/chat/completions
  • Models endpoint at /v1/models for capability discovery
  • Public status endpoint at /v1/stats for live node health

Advertised performance characteristics (TTFT typical ~64ms, throughput ~62 tok/s, context 28K, precision FP8) reflect the latest published benchmark measured under controlled benchmark conditions on a single RTX 5090 · 32 GB dedicated node. Live numbers may vary; see System Status for real-time measurements.

03 Acceptable Use

You agree not to use the Service to:

  • Generate content that is unlawful, harmful, harassing, or infringing
  • Generate, distribute, or solicit child sexual abuse material (CSAM) or any sexual content involving minors
  • Attempt to evade economic sanctions or use the Service from jurisdictions where it is prohibited
  • Bypass safety measures of upstream model providers
  • Attempt to exfiltrate model weights, system prompts, or KV cache contents
  • Conduct load tests or stress probes that exceed the published concurrency ceiling (64 concurrent requests per API key). Excess load will receive HTTP 429.
  • Resell access without prior written agreement

We report suspected CSAM and cooperate with law enforcement as required by applicable law. Abuse reports may be sent to legal@gwmmai.com.

We may rate-limit or suspend API keys that violate this section. Operational metadata (token counts, timestamps, error codes) is retained up to 90 days as documented in the Privacy Policy.

max_concurrency = 64 · first_token_timeout = 20 s · context_window = 28K

04 Billing

GWMM operates two billing paths:

  • GWMM direct wallet (Stripe). You top up a prepaid wallet through Stripe Checkout. Usage is deducted from your wallet balance in real time, priced per token at the rates listed on Models & Pricing. Wallet balances do not expire. We do not offer self-service refunds; refund requests may be submitted to help@gwmmai.com and will be reviewed on a case-by-case basis.
  • OpenRouter route. When you access GWMM through OpenRouter, billing is handled entirely by OpenRouter. Token counts are reported by the GWMM gateway to OpenRouter for invoicing. Disputes, refunds, and invoice history are managed through the OpenRouter dashboard.

Pricing is denominated in US dollars per million tokens, with separate input and output rates as published.

05 Disclaimers & Liability

# TODO(legal): Placeholder for legal review. The following language is a working draft and has not been approved by counsel. Specific liability caps, indemnification clauses, and warranty disclaimers will be inserted before these Terms are presented to a customer.

The Service is provided on an “as-is” and “as-available” basis. To the maximum extent permitted by applicable law, GWMM disclaims all warranties, express or implied, including warranties of merchantability, fitness for a particular purpose, and non-infringement. GWMM does not warrant that the Service will be uninterrupted, error-free, or that model outputs will be accurate.

You are solely responsible for your use of model outputs, including any decisions made or actions taken in reliance on those outputs. GWMM is not liable for any indirect, incidental, special, consequential, or punitive damages arising out of or related to your use of the Service.

06 Changes

We may update these Terms from time to time. Material changes will be announced on the System Status page and via the OpenRouter provider changelog. Continued use of the Service after the effective date of a change constitutes acceptance of the revised Terms.

07 Contact

Questions about these Terms or our service commitments? Reach the team directly.

help@gwmmai.com · legal@gwmmai.com (privacy / compliance)