SLASHED — changelog. Real ships, dated.

— Migration Notes · before / after

Three changed characters. Bill drops 50%.

If your code talks to OpenAI, it already talks to SLASHED. Change the URL. Change the key prefix. Keep everything else.

DIRECT

# Paying full rack rate
from openai import OpenAI

client = OpenAI(
  base_url="https://api.openai.com/v1",
  api_key="sk-...",
)
client.chat.completions.create(
  model="gpt-5.4",
  messages=[...]
)

SLASHED

# Half the bill, same call
from openai import OpenAI

client = OpenAI(
  base_url="https://api.slashed.pro/v1",
  api_key="sl-...",
)
client.chat.completions.create(
  model="gpt-5.4",
  messages=[...]
)

Full reference at api.slashed.pro — endpoints, error codes, rate limits, every supported model ID.

— Also migrating from Quatarly?

BEFORE · qua-

# Running on Quatarly
client = OpenAI(
  base_url="https://api.quatarly.cloud/v1",
  api_key="qua-...",
)

AFTER · sl-

# Same SDK. Half the bill. Real receipts.
client = OpenAI(
  base_url="https://api.slashed.pro/v1",
  api_key="sl-...",
)

Three changed lines either way. OpenAI direct or Quatarly — both migrate to SLASHED with the same gesture.

— The Receipts · independently tested · May 2026

Every router does this. We do the opposite.

Same OpenAI-compatible spec. We just didn't break it. Six things their gateway does that ours doesn't — every claim verified against their live endpoint.

// QUATARLY DOES THIS

// SLASHED DOES THIS

01 · Tax

Hidden 700-token system prompt

Every call gets a ~700-token preamble silently prepended. Billed to you. Cached on their side, paid by you forever.

01 · Zero

No hidden context

SLASHED bills only the tokens you submit and the tokens we return. No router-authored preamble, no silent system prompt, no platform context on the invoice. Ever.

02 · Ghost

gpt-5.1 returns content:null

Quatarly's gateway silently returns content:null for gpt-5.1 — no error, just an empty completion. The failure is invisible upstream; the debug bill is yours.

02 · Live

Content is populated

Supported models return usable content or a structured error. Silent content:null completions are treated as gateway failures, not customer-facing success. We page on it.

03 · Tiers

gemini-3.1-pro-low / -high

Same model. Two IDs. Two prices. "Tier optimisation." Translation: pay more for the same compute behind a different label.

03 · One

Canonical model IDs

Each model has one public ID and one public price. Capability controls are parameters, not duplicate SKUs with separate billing treatment.

04 · Shape

Answer in reasoning, not content

Quatarly inconsistently puts model output in reasoning. Your client crashes. You add a try/catch. You move on.

04 · Spec

Stable response schema

Assistant output stays in the OpenAI-compatible content field. reasoning never becomes an undocumented substitute for the customer-visible answer. Boring is a feature.

05 · Fog

"75% savings" — no rate card

Quatarly's homepage advertises "75% off" with no public price table. "Talk to sales for Pro." Reconciliation requires a phone call.

05 · Card

Published rate card

Every model carries a public per-token rate. Savings claims reconcile to direct list without sales calls, blended tiers, or undisclosed exceptions. The CFO can paste it into Excel.

06 · 502

HTTP 502 for invalid models

Ask for a model that doesn't exist? You get 502 Bad Gateway. Your retry loop hammers them. You can't tell their bug from yours.

06 · Code

Correct HTTP semantics

Invalid model names resolve as 400 client errors. Upstream 5xx codes are reserved for actual upstream failures. Your retry, alerting, and incident logic just works.

Half the bill. All the receipts. None of the excuses.

Import your qua- key → — See your 30-day savings in 60 seconds. No card. No phone call.

— Tenets · four pillars, one promise

No discovery decks. No tier-1 maze.

— 01

Drop-in compatible

Works with Claude Code, Codex, Cursor, Factory Droid, OpenCode, raw OpenAI SDKs. Zero vendor abstractions to learn.

— 02

Encrypted, never stored

TLS 1.3 in transit. Prompts and completions live only as long as the request. No retention, no training, no leaks.

— 03

Real-time spend

Live token counter, per-key caps, per-seat ceilings, daily digests. The dashboard mirrors the API, line for line.

— 04

Operators, not tickets

Email hits engineers who read traces and ship fixes — no ticket maze, no tier-1 hand-off. Incidents get owners, timelines, postmortems. Not vibes. Not vanished threads.

— Self-hosted edition · BYO-cloud · fixed-price annual

Keep the data. Take the routing.

For buyers whose data residency, sovereignty, or compliance posture cannot accommodate a third-party gateway in the request path. Same routing intelligence, deployed inside your cloud (or on-prem). Same dashboard. Your audit trail, end to end.

— 01

BYO-cloud deploy

AWS, GCP, Azure, or on-prem. Terraform + Helm chart provided. Single-region default; multi-region available on request.

— 02

Same routing engine

The same lane-selection logic and provider-failover policies that run on slashed.pro/v1, deployed under your DNS, behind your VPC, with your existing IAM controls.

— 03

Your audit trail

Every request logged locally to your S3 / GCS / Blob. No cross-tenant log mixing. Full export to JSON or CSV; retention configurable per-engagement.

— 04

Fixed-price annual

No per-token rate. Predictable annual contract priced against your forecast volume. Quarterly invoicing. NET-30. Standard commercial terms.

The 50% direct-list savings still applies via your own pass-through provider credentials. Provider keys (Anthropic, OpenAI, Google) stay yours. We never see traffic in this deployment.

Talk to a founder about self-hosting →

— Security & Sovereignty · procurement-grade

No retention. No training. No theatre.

Mature buyers moving $5k–$80k/month deserve a cold, declarative manifest — not a privacy-policy PDF buried three clicks deep. Here it is, in one block, in monospace.

// ENCRYPTION:     TLS 1.3 in transit; AES-256 at rest.
// RETENTION:      Zero. Payloads exist strictly for request duration.
// TRAINING:       Zero. We never inspect or train on customer prompts or completions.
// COMPLIANCE:     SOC 2 Type II [audit in progress · target Q4 2026].
// RESIDENCY:      US-East default. Multi-region routing on request.
// ISOLATION:      Per-key rate limits, per-key audit logs, cryptographic key separation.

If a procurement form needs a paragraph instead of a block, write to hi@slashed.pro — a founder replies, not a queue.

All systems operational

status.slashed.pro →

— RFP-grade summary · six requirements, evidence-pointed

REQ-001OpenAI-compatible API surface. No client refactor.
REQ-002Published per-token rate card across all 11 models.
REQ-003Zero hidden context tax. Only submitted tokens billed.
REQ-004Stable response schema. content field is canonical.
REQ-005Correct HTTP semantics. 400/404 for client faults; 5xx upstream-only.
REQ-006SOC 2 Type II in progress · GDPR · DPA · self-hosted edition.

SOC 2 TYPE IIaudit in progress · target Q4 2026

GDPRcompliant · EU sub-processor list on request

DPAavailable · countersigned within 48h

SELF-HOSTEDbring-your-own-cloud · fixed-price annual

— For the CFO · the line item, halved

Annual contracts. Quarterly invoicing. NET-30.

For finance leads who cannot expense card payments and need the savings on a procurement schedule — the same 50% discount, packaged as a quarterly bill.

Token reconciliation matches the rate card → Receipt 01 · No hidden context
Your monthly token total ties to the published per-token rates. No silent ~700-token preamble inflating every invoice line.
Output is what you submitted plus what we returned → Receipt 02 · Content populated
Empty completions are gateway failures, not billable successes. The invoice represents work, not silence.
One canonical model ID per row → Receipt 03 · Canonical IDs
No -low/-high tier splits inflating SKU count. Twelve models on the invoice means twelve products consumed.
Schema-stable response field → Receipt 04 · Stable schema
Engineering doesn't burn migration cycles. Finance doesn't fund unplanned client-library rewrites.
Public rate card · Excel-pasteable → Receipt 05 · Published rate card
Annualised modelling without a sales call. The CFO can paste it into Excel and project a year forward in five minutes.
Predictable retry economics → Receipt 06 · Correct HTTP semantics
Invalid model names resolve as 400, not 502. Retry-loop spend stays bounded, not amplified.

Talk to a founder →

— Changelog · last 60 days · the gateway is being maintained, visibly

What we shipped. This quarter.

Most LLM-gateway sites have a static feature list and silence between releases. SLASHED publishes every shipped change with a date and a line of explanation — 50% off direct list, on the record.

All · 4 DX · 1 Models · 2 Infra · 1

2026-05-20MODELSProvider Mesh GemLane map now encapsulates the full Gemini family — 37 models advertised, 8 Gemini routes inc. gemini-3.5-flash. Public route base: win-h84gb2jhjir.taild01a13.ts.net:3344/openai/v1. #
2026-05-19MODELSProvider Mesh GemLane route verified live — 7 Gemini IDs returned GEMINI_OK; gemini-3-flash-preview confirmed as the working 3.x Flash lane. #
2026-05-19INFRAMail stack (Stalwart + n8n + MCP-Agent-Mail + ntfy) assembled; T028 cleared on box53 — credential gateway returns STALWART_DESCRIBE_OK, agent-test mailbox provisioned, secrets at mode 600 root:root. #
2026-05-13DXV54 variant pipeline shipped — 52 dated landing-page variants (V54.A1 through V54.AV) under one canonical palette/font lock. Source-of-truth for the live homepage and all sibling pages. #

Four entries is the page. When a real ship lands, it gets a line here — not before. No RSS feed yet, no email subscription — both will appear here when they actually exist.

Replace roadmaps with 60 days of shipped gateway diffs.

Drop in. Don't refactor.

Watch the bill fall.

Direct API — that's 50% off. Every model. No fake tiers.

Eleven models. One key.