DevPass

One AI coding subscription. Every model.
by @llmgateway

[DevPass @LLMDevPass]

Gemini 3.1 Flash Image ("Nano Banana 2") is live on LLM Gateway: fast native image gen, 128K context, $0.50/M in / $3/M out.
https://t.co/IdWO4JRv61

[Smakosh @smakosh]

Your usage comparisons don’t have to follow the calendar anymore. 📊

In @llmgateway, you can now choose the exact start date for week-over-week or month-over-month comparisons, making changes in token usage and cost much easier to spot.

Should be live soon

[DevPass @LLMDevPass]

Unpopular opinion: a coding plan should print its cap on the card.

DevPass does. Every dollar buys $3 of model usage at provider list rates, and the weekly frontier allowance is 12%, 15% or 18% of credits by tier.

No "unlimited" that quietly isn't.

https://t.co/zfkVQH6L8K

[DevPass @LLMDevPass]

Muse Spark 1.3 is live on LLM Gateway: Meta's agentic reasoning model for long-horizon coding, 1M context, $1.25/M in / $4.25/M out.
https://t.co/fAeWi5EnIv

[Smakosh @smakosh]

Muse Spark 1.3 and 1.3 contributor are live on @llmgateway & @LLMDevPass

- https://t.co/gwL4OVXKW2
- https://t.co/k9oOnNxLTa

[DevPass @LLMDevPass]

Recomputed every few minutes, and today's leader usually isn't tomorrow's: https://t.co/WEzaMgwsh4

[DevPass @LLMDevPass]

Last 24h on the gateway: 22.9B tokens, 384.8K requests, 110 models ranked.

#1 DeepSeek V4 Pro holds 14.8% of that. That's what leading looks like now.

Gemini 3.8 Flash, up 298.2%, already routes more requests than any model on the board.

[DevPass @LLMDevPass]

The top three sit inside two points of each other. It reshuffles hourly: https://t.co/WEzaMgwsh4

[DevPass @LLMDevPass]

Last 24h on the gateway: 20.1B tokens, 381.5K requests, 118 models ranked.

#1 and #2 finished 50.2M tokens apart — a quarter of one percent of the board.

GLM-5.3 Flash 3.3B, down 23.2%. DeepSeek V4 Flash 3.2B, down 7.2%.

Nobody owns the top spot.

[DevPass @LLMDevPass]

Gemini 3.8 Flash is live on LLM Gateway: 1M context, reasoning built in, $0.75/M in / $3.75/M out.
https://t.co/ACIQNOeg03

[DevPass @LLMDevPass]

Claude Fable 5.1 is live on LLM Gateway: Anthropic's flagship for long-horizon agentic work. 1M context, adaptive reasoning, $10/M in / $50/M out, cache reads at $0.25/M.
https://t.co/VVx86Fzgfk

[DevPass @LLMDevPass]

Most coding plans meter you in credits, messages or "fast requests" — units only the vendor can price.

DevPass meters in dollars at provider list rates: $29 → $87, $79 → $237, $179 → $537.

Same allowance in Claude Code, Cline, Cursor, OpenCode.

https://t.co/zfkVQH6L8K

[DevPass @LLMDevPass]

The month underneath the #1 spot: GPT-5.6 Terra +657%, GPT-5.6 Luna +541% into #4, GLM-5.2 +102%.

Gemini 3.7 Flash logged its first tokens on Aug 13 and still closed the month at #5 with 39.2B.

Half a month was enough.

https://t.co/WEzaMgwsh4

[DevPass @LLMDevPass]

30 days on the gateway: 499.9B tokens, 15.2M requests, 206 models ranked.

DeepSeek V4 Flash took the month at 118.0B — 23.6% of every token we routed, up 710% on the prior 30 days.

One model, a quarter of the board.

[Smakosh @smakosh]

lol what, are you that far behind?

1. Get @LLMDevPass for x3 usage
2. Use GLM 5.3 and GLM 5.3 Flash (50% off until 9th September)

- https://t.co/wD0qJqeTgK
- https://t.co/omt4eLpT55

Works with any coding agent

[DevPass @LLMDevPass]

GLM-5.3 grew 429.7% this week. GLM-5.2, the model it replaced, fell 61.5%.

Same lab, one generation apart, seven days.

https://t.co/WEzaMgwsh4

[DevPass @LLMDevPass]

7 days on the gateway: 119.6B tokens, 3.9M requests, 163 models ranked.

#1 was DeepSeek V4 Flash at 17.3B — down 22.8% from last week and still on top.

Nothing grew like GLM-5.3: +429.7%.

The board doesn't reward incumbency.

[DevPass @LLMDevPass]

102 models, reranked every few minutes. Live board: https://t.co/WEzaMgwsh4

[DevPass @LLMDevPass]

Your default model has a shelf life.

Last 24h on the gateway: 19.2B tokens, 560.1K requests, 102 models ranked.

GLM-5.3 Flash took #1 with 17.4% of all tokens, up 148.1%.
Last generation's GLM-5.2: down 49.5%.

Hardcode one model and you're pinned to last month's board.

[Smakosh @smakosh]

@OpenAI @CARLOSZIEGLER 1- get a sub on @LLMDevPass
2- follow this guide https://t.co/FiI6wPJqyf