LLM Gateway is a hosted model gateway from LLMGateway, Inc. (GitHub org theopenco) that routes 250+ models from 40+ providers behind one OpenAI-compatible endpoint at provider list prices with a 5% fee on credit top-ups, and sells DevPass, a flat-rate multi-model coding subscription at $29, $79, and $179 per month. It runs the OpenRouter playbook at a lower fee and bolts on the flat coding plan OpenRouter never built, which makes it the only member here selling percentage and flat-rate billing side by side.
What it is #
A hosted gateway with a self-host option: pay per token at each provider’s own rates with no token markup, a 5% platform fee when you buy credits, and BYOK routing with no platform fee at all.
The repo ships under AGPLv3 with a commercial carve-out for the ee/ directory, so self-hosting is free and the enterprise features are the licensed part.
DevPass claims every $1 converts into $2 of model usage metered at provider rates, works from DevPass Code, Claude Code, OpenCode, Empryo, SoulForge, or any OpenAI-compatible tool, and gates premium models (roughly $5+ per 1M input) behind a weekly fair-use share of credits.
The company reports SOC 2 Type II, a 30-day enterprise pilot with a 99.9% SLA, and site counters of 1T+ tokens and 80M+ requests routed, as of 2026-10-04.
Status #
Active and young: the repo was created 2025-04-12 and shows 1,687 stars and 193 forks as of 2026-10-16, pushed 2026-10-22 (as of 2026-10-23).
DevPass shipped across Q2 2026 with annual billing and integration guides, and the Q2 roundup puts deepseek-v4-pro at the top of its quarterly token table. A Hacker News search this run returned no stories about the product, so the footprint is SEO and docs, not community debate. On 2026-10-04 the DevPass page announced its first devaluation: from October 15, 2026 the monthly allowance falls from 3x to 2x the plan price. The cut landed on schedule: at the 2026-10-12 and 2026-10-14 re-checks every plan card still read 3x usage, and on 2026-10-16 every card reads 2x, with the page’s Plan-changes section stating the new daily caps, tighter premium weekly caps, and updated Reset Pass benefits apply from October 15 even inside existing billing cycles.
Strengths #
- The fee undercuts OpenRouter’s 5.5% credit fee and drops the $0.80 minimum entirely, while BYOK routing costs nothing at all.
- DevPass is the only flat-rate plan here spanning every frontier family (Claude, GPT-5, Gemini) plus the open-weight coders under one key, with a 14-day self-serve refund on the first month (under 10% of the allowance used).
- Spend controls, prompt caching, per-project ceilings, and provider-compliance routing (SOC 2, ISO 27001, and GDPR providers only) are documented on the public pricing page.
- Self-hosting exists if the hosted fee is unacceptable, though enterprise features need a license.
Cautions #
- The community footprint is thin: no Hacker News stories, and the company’s “best AI coding plans” post ranks DevPass first against Claude Max, OpenCode Go, and the GLM plan, which is marketing, not measurement.
- DevPass value fell on October 15, 2026, from 3x to 2x the plan price at renewal, the first cut since launch, and the same change tightened the daily caps, the weekly premium caps, and the Reset Pass prices inside existing billing cycles; the weekly frontier fair-use share means heavy premium-model users exhaust headroom before the allowance runs out.
- The 5% is charged per top-up, so the small-purchase math that punishes OpenRouter’s $0.80 minimum applies here too for frequent small reloads.
- A young repo (1,687 stars as of 2026-10-16) is early to hold custody of every provider key you own.
Pricing #
PAYG: free tier with 3 rate-limited free models, provider list prices, 5% fee on credit top-ups, BYOK free, non-US cards add 1.5%, and optional full payload retention costs $0.01 per 1M tokens, as of 2026-10-07 and re-verified 2026-10-14 (top-ups run $10 minimum to $5,000 maximum per the pricing page). DevPass (as of 2026-10-16, the first devaluation landed): Lite $29/month (about $58 of usage), Pro $79/month (about $158), Max $179/month (about $358), worth 2x the plan price since October 15, 2026 (3x before), 14-day self-serve refund on the first month with under 10% of the allowance used, daily pacing of 8%/9%/10% of credits per rolling 24 hours, and weekly frontier fair-use at 10% of credits on Lite, 12% on Pro, and 15% on Max; the cards publish Reset Passes (instant premium-allowance resets) at $5 each on Lite, 1 per month on Pro with extras at $15, and 2 per month on Max with extras at $45. Lounge chat plans: Starter $9/month, Plus $19/month, Pro $49/month. Enterprise: custom, 30-day pilot, SAML SSO and SCIM, 99.9% SLA.
Price history #
| Date | Plan | Change | Source |
|---|---|---|---|
| 2026-10-04 | PAYG / DevPass | Baseline: PAYG at provider rates with 5% on credit top-ups and free BYOK; DevPass Lite $29, Pro $79, Max $179 per month at 3x usage value | |
| 2026-10-15 | DevPass | Announced: included usage falls from 3x to 2x the plan price at first renewal on or after this date (new subscriptions start at 2x), prices unchanged | |
| 2026-10-15 | DevPass | Landed as announced (observed 2026-10-16): every card reads 2x, and the same change tightened daily pacing to 8%/9%/10% of credits per day, weekly frontier fair-use to 10%/12%/15%, Reset Pass prices to $5/$15 extras/$45 extras, and the refund threshold to under 10% of the allowance |
Compared to #
OpenRouter is the same pay-as-you-go gateway idea with the deeper catalog and track record at 5.5% per credit purchase; choose LLM Gateway for the lower fee or for DevPass, OpenRouter for maturity and reach. Requesty meters 5% of upstream spend with EU residency; LLM Gateway’s 5% sits on top-ups instead, which is kinder to steady spend but carries no residency story. LiteLLM is the self-hosted zero-fee incumbent; pick it when keys cannot leave your perimeter, LLM Gateway when you want someone else to run the gateway.
Bottom line #
Recommended for multi-model developers who want hosted gateway routing at a fee matching or beating every competitor here, and for flat-rate coding across all frontier families at DevPass prices. Not for teams that need community scrutiny and a battle-tested gateway, or for premium-model-heavy users who would blow through the fair-use caps. My disagreeable claim: DevPass at 2x buys less leverage than OpenCode Go’s 6x on open models, so its whole case rests on needing frontier families under one flat bill.
Changes #
- 2026-10-04 - Created when the new-entrant scan surfaced DevPass marketed against this category’s members; added with the announced October 15 usage cut already on the price table.
- 2026-10-06 - Converted the Compared-to cross-references from plain-text paths into working links.
- 2026-10-06 - DevPass re-check ahead of the October 15 change: the notice now also adds daily caps and tighter premium weekly caps applying inside existing billing cycles, and the cards publish the frontier fair-use share per tier (Lite 12%, Pro 15%, Max 18% of credits); plan prices and the 3x-to-2x cut re-verified unchanged.
- 2026-10-07 - Added the theopenco/llmgateway star history chart to the Status section.
- 2026-10-07 - Corrected the license claim: the LICENSE file and the pricing page now state AGPLv3 with the commercial
ee/carve-out, where this note had recorded a custom non-OSI license from the 2026-10-04 read. - 2026-10-07 - Pricing-page completeness: the Lounge lineup named precisely (Starter $9, Plus $19, Pro $49) and the DevPass cards now publish Reset Pass pricing ($9 Lite extras, 1 per month Pro with $29 extras, 2 per month Max with $79 extras); PAYG and DevPass plan prices and the October 15 3x-to-2x change notice re-verified unchanged.
- 2026-10-11 - DevPass re-check four days before the announced cut: the October 15 3x-to-2x devaluation is still pending with every card reading 3x usage, the refund policy is detailed as first-month-only and void past 20% of the allowance used, and the named client list moved to DevPass Code, Claude Code, OpenCode, Empryo, and SoulForge; plan prices re-verified unchanged.
- 2026-10-16 - The devaluation landed on schedule: every card reads 2x usage (Lite $58, Pro $158, Max $358), and the same October 15 change tightened daily pacing to 8%/9%/10% of credits per day, weekly frontier fair-use to 10%/12%/15% (from 12%/15%/18%), Reset Pass prices to $5/$15 extras/$45 extras (from $9/$29/$79), and the refund threshold to under 10% of the allowance (from 20%); price history row appended and the usage-claim sentence moved to $2 per $1.
See also #
- OpenRouter - the gateway whose fee model this one undercuts.
- Requesty - the other 5% gateway, metering spend instead of top-ups.
- OpenCode Go - the flat-rate open-model alternative DevPass is priced against.
- LiteLLM - the self-hosted zero-fee counterpoint.
- Model provider feature matrix - where these gateways meet their upstream vendors.
References #
https://llmgateway.io/ - 250+ models from 40+ providers, 5% on top-ups, BYOK free, SOC 2 Type II, 1T+ tokens routed, self-host with licensed enterprise features (200, fetched 2026-10-04).
https://llmgateway.io/pricing - free tier, 5% platform fee, BYOK included, retention pricing, plan comparison table (200, fetched 2026-10-04).
https://llmgateway.io/pricing - free tier, 5% platform fee, BYOK included, retention pricing, plan comparison table (200, fetched 2026-10-04; re-fetched 2026-10-14 with the $10-to-$5,000 top-up bounds and the DevPass/Lounge lineups unchanged).
https://devpass.llmgateway.io/ - DevPass $29/$79/$179, $1-to-$3 usage claim, frontier fair-use, 14-day refund, supported agents (200, fetched 2026-10-04).
https://devpass.llmgateway.io/pricing - per-plan usage values and the October 15, 2026 3x-to-2x change notice (200, fetched 2026-10-04, re-fetched 2026-10-11, 2026-10-12, and 2026-10-14 with the cut still pending and every card reading 3x; fetched 2026-10-16 with the cut landed, every card reading 2x, and the tightened daily pacing, fair-use, Reset Pass, and refund terms live)
https://llmgateway.io/blog/q2-2026-roundup - DevPass Q2 launch detail, SOC 2 Type II, top-routed models by tokens (200, fetched 2026-10-04).
https://api.github.com/repos/theopenco/llmgateway - 1,674 stars, 191 forks, created 2025-04-12, pushed 2026-10-07 (200, fetched 2026-10-07; re-fetched 2026-10-14 at 1,685 stars, 193 forks, pushed 2026-10-13; re-fetched 2026-10-18, pushed 2026-10-18; re-fetched 2026-10-19, pushed 2026-10-19, v1.21.0 still latest; re-fetched 2026-10-21, pushed 2026-10-20, v1.22.0 released 2026-10-19; re-fetched 2026-10-23, pushed 2026-10-22, v1.22.0 still latest).
https://raw.githubusercontent.com/theopenco/llmgateway/main/LICENSE - AGPLv3 grant with the ee/ commercial carve-out (200, re-fetched 2026-10-07; the 2026-10-04 read had recorded it as a custom license).
https://llmgateway.io/blog/best-ai-coding-plans - the company’s self-ranking comparison against Claude Max, OpenCode Go, and the GLM plan, cited as marketing positioning rather than measurement (200, fetched 2026-10-04).