between this and coming across a chart from OpenRouter that showed US model token use going flat over the last few months, I do think we’re reaching the limits of how much people are willing to spend on these things
Yes and replit just dumped Anthropic entirely due to inflexibility over discounts. OpenAI gave deep discounts to get them. OpenAI is also planning to start limiting paid accounts’ token burn in December so I can’t imagine these are permanent discounts either
 bsky.app/profile/sbis...
 bsky.app/profile/sbis...
Replit is one of the first companies to fully pivot to an AI garbage product, right?
November 2, 2026 at 7:36 AM
between this and coming across a chart from OpenRouter that showed US model token use going flat over the last few months, I do think we’re reaching the limits of how much people are willing to spend on these things
hermes + an openrouter api key is the way to go if you want to be a lazy piece of shit like me imho
October 31, 2026 at 3:01 PM
hermes + an openrouter api key is the way to go if you want to be a lazy piece of shit like me imho
because I don't use LLMs enough to justify paying for subscriptions goobers. I put 30 dollars in my openrouter account months ago and I still have like 25 dollars
November 1, 2026 at 5:14 AM
because I don't use LLMs enough to justify paying for subscriptions goobers. I put 30 dollars in my openrouter account months ago and I still have like 25 dollars
OpenRouter shows us it's false!
October 30, 2026 at 8:55 PM
OpenRouter shows us it's false!
for deepseek specifically im getting api prices wayyyy below theirs tho from a provider that isnt listed on openrouter i think the options exist
November 1, 2026 at 3:41 PM
for deepseek specifically im getting api prices wayyyy below theirs tho from a provider that isnt listed on openrouter i think the options exist
Whenever my credit gets low I plan on switching over to deepseek directly! Also openrouter has byok so I can just continue to use my openrouter key for it (it's free until you spend over 25k a month)
November 1, 2026 at 5:18 AM
Whenever my credit gets low I plan on switching over to deepseek directly! Also openrouter has byok so I can just continue to use my openrouter key for it (it's free until you spend over 25k a month)
Dovrei riprenderle ma credo prenderò altri crediti per OpenRouter
November 2, 2026 at 9:31 AM
Dovrei riprenderle ma credo prenderò altri crediti per OpenRouter
Ага, и меня забанили. Благо я с Клодом написал собственного ИИ-агента, а доступ к моделям получаю через OpenRouter.
November 2, 2026 at 3:25 PM
Ага, и меня забанили. Благо я с Клодом написал собственного ИИ-агента, а доступ к моделям получаю через OpenRouter.
i'm very sympathetic to this view but after some chat-math, i think the best hope is openrouter-style aggregration services. the problem to be solved is diversification and flexibility, not custody.
My preference for the future of inference looks a lot more like collectives or shared compute.
For example, I'd love to go in on a Kimi node with a large group of people, and allow us to pay for GPU time and get unlimited (if delayed) tokens.
For example, I'd love to go in on a Kimi node with a large group of people, and allow us to pay for GPU time and get unlimited (if delayed) tokens.
November 1, 2026 at 12:38 AM
i'm very sympathetic to this view but after some chat-math, i think the best hope is openrouter-style aggregration services. the problem to be solved is diversification and flexibility, not custody.
lmao T3 Code is already on track for 5.5 quadrillion tokens a year
DMs are open btw
DMs are open btw
November 2, 2026 at 10:57 AM
lmao T3 Code is already on track for 5.5 quadrillion tokens a year
DMs are open btw
DMs are open btw
Swift Language Models brings other providers' models into Apple's native Foundation Models API. Keep LanguageModelSession, tools and @Generable, and switch models by changing only model:
First provider: OpenRouter, with hundreds of models. More coming.
github.com/dioKaratzas/swift-language-models
First provider: OpenRouter, with hundreds of models. More coming.
github.com/dioKaratzas/swift-language-models
GitHub - dioKaratzas/swift-language-models: A unified Swift API for Apple’s Foundation Models-style experience across cloud and local LLM providers. Bring OpenRouter, Ollama, Claude, GPT, Gemini and m...
A unified Swift API for Apple’s Foundation Models-style experience across cloud and local LLM providers. Bring OpenRouter, Ollama, Claude, GPT, Gemini and more into a consistent LanguageModelSessio...
github.com
October 31, 2026 at 4:44 PM
Swift Language Models brings other providers' models into Apple's native Foundation Models API. Keep LanguageModelSession, tools and @Generable, and switch models by changing only model:
First provider: OpenRouter, with hundreds of models. More coming.
github.com/dioKaratzas/swift-language-models
First provider: OpenRouter, with hundreds of models. More coming.
github.com/dioKaratzas/swift-language-models
Stealth Model Alert
OpenRouter:
• stealth/crouton-alpha 🛑
Opencode Zen:
• crouton-alpha-free 🛑
+2 added · 0 removed
#freeLLMs #opensourceai
OpenRouter:
• stealth/crouton-alpha 🛑
Opencode Zen:
• crouton-alpha-free 🛑
+2 added · 0 removed
#freeLLMs #opensourceai
November 3, 2026 at 9:00 PM
Stealth Model Alert
OpenRouter:
• stealth/crouton-alpha 🛑
Opencode Zen:
• crouton-alpha-free 🛑
+2 added · 0 removed
#freeLLMs #opensourceai
OpenRouter:
• stealth/crouton-alpha 🛑
Opencode Zen:
• crouton-alpha-free 🛑
+2 added · 0 removed
#freeLLMs #opensourceai
also, unless you configure it, it will only work for your own account to prevent it from getting super expensive & heavy to run -> bsky.app/profile/j4ck...
for me, it's using about 350MB RAM with 3 feeds being served just for me :)
(using hosted LLMs via OpenRouter)
for me, it's using about 350MB RAM with 3 feeds being served just for me :)
(using hosted LLMs via OpenRouter)
November 3, 2026 at 6:27 PM
also, unless you configure it, it will only work for your own account to prevent it from getting super expensive & heavy to run -> bsky.app/profile/j4ck...
for me, it's using about 350MB RAM with 3 feeds being served just for me :)
(using hosted LLMs via OpenRouter)
for me, it's using about 350MB RAM with 3 feeds being served just for me :)
(using hosted LLMs via OpenRouter)
you can byok for deepseek on openrouter and it is free for the first 25k of spend and i dont have 25k so i will probably just do that. thanks for the advice oomfie
October 31, 2026 at 3:38 PM
you can byok for deepseek on openrouter and it is free for the first 25k of spend and i dont have 25k so i will probably just do that. thanks for the advice oomfie
OpenRouter hit an annualized run rate of over 10 quadrillion tokens after processing about 194T in a single October week—more than its prior 39-month cumulative output.
Autonomous agents burned about 7x more tokens than humans by late September, fueling that roughly 5% take rate.
Autonomous agents burned about 7x more tokens than humans by late September, fueling that roughly 5% take rate.
OpenRouter Passes 10 Quadrillion Annual Token Run Rate as AI Agents Pass 7x Human Usage
OpenRouter surpassed an annualized run rate of over 10 quadrillion tokens after processing approximately 194 trillion tokens in the week ending October 26, 2026.
zubiqo.com
November 2, 2026 at 8:59 PM
OpenRouter hit an annualized run rate of over 10 quadrillion tokens after processing about 194T in a single October week—more than its prior 39-month cumulative output.
Autonomous agents burned about 7x more tokens than humans by late September, fueling that roughly 5% take rate.
Autonomous agents burned about 7x more tokens than humans by late September, fueling that roughly 5% take rate.
OpenRouter tops: SpaceXAI’s Grok 4.20 (multi‑agent) & Grok 4.20 with Auto Router for huge context. Cheapest: Z.ai’s GLM 5.2, OpenAI’s gpt‑oss‑120b & inclusionAI’s Ling 3.0 Flash.
#news #ai #artificialintelligence #technology #grok
#news #ai #artificialintelligence #technology #grok
November 1, 2026 at 1:40 PM
OpenRouter tops: SpaceXAI’s Grok 4.20 (multi‑agent) & Grok 4.20 with Auto Router for huge context. Cheapest: Z.ai’s GLM 5.2, OpenAI’s gpt‑oss‑120b & inclusionAI’s Ling 3.0 Flash.
#news #ai #artificialintelligence #technology #grok
#news #ai #artificialintelligence #technology #grok
Really what AI is is a force for centralising power over cognition in the hands of one company. Ignore this screenshot i can’t figure out how to delete it sorry
October 30, 2026 at 10:25 PM
Really what AI is is a force for centralising power over cognition in the hands of one company. Ignore this screenshot i can’t figure out how to delete it sorry
for some models it's a license thing and we should be angry at intellectual property (as usual) but for many it's not and. idk how the fuck the industry is so oligopoly-like when the cost of entry is also so low (other than that you need reputation and shit to get into the markets like openrouter)
November 1, 2026 at 3:36 PM
for some models it's a license thing and we should be angry at intellectual property (as usual) but for many it's not and. idk how the fuck the industry is so oligopoly-like when the cost of entry is also so low (other than that you need reputation and shit to get into the markets like openrouter)
yeah roleplay + conversation are non-zero parts of the openrouter 'general' leaderboard openrouter.ai/rankings#lea..., but given that it's fewer tokens and smaller models, there have to be a *ton* of people using it this way
October 30, 2026 at 5:32 PM
yeah roleplay + conversation are non-zero parts of the openrouter 'general' leaderboard openrouter.ai/rankings#lea..., but given that it's fewer tokens and smaller models, there have to be a *ton* of people using it this way
📉 Claude Opus 5.5: 55% off vs OpenRouter
• $3.60 per 1M tokens right now
• OpenRouter charges $8.00
• Up 25.5% in the last 24h
Which model should we chart next?
https://qubax.ai/market/claude-opus-5.5
• $3.60 per 1M tokens right now
• OpenRouter charges $8.00
• Up 25.5% in the last 24h
Which model should we chart next?
https://qubax.ai/market/claude-opus-5.5
November 3, 2026 at 6:00 PM
📉 Claude Opus 5.5: 55% off vs OpenRouter
• $3.60 per 1M tokens right now
• OpenRouter charges $8.00
• Up 25.5% in the last 24h
Which model should we chart next?
https://qubax.ai/market/claude-opus-5.5
• $3.60 per 1M tokens right now
• OpenRouter charges $8.00
• Up 25.5% in the last 24h
Which model should we chart next?
https://qubax.ai/market/claude-opus-5.5
📉 Kimi K3.1 Pro: 55% off vs OpenRouter
• $1.78 per 1M tokens right now
• OpenRouter charges $4.00
• Down 1.0% in the last 24h
Which model should we chart next?
https://qubax.ai/market/kimi-k3.1-pro
• $1.78 per 1M tokens right now
• OpenRouter charges $4.00
• Down 1.0% in the last 24h
Which model should we chart next?
https://qubax.ai/market/kimi-k3.1-pro
November 2, 2026 at 6:00 PM
📉 Kimi K3.1 Pro: 55% off vs OpenRouter
• $1.78 per 1M tokens right now
• OpenRouter charges $4.00
• Down 1.0% in the last 24h
Which model should we chart next?
https://qubax.ai/market/kimi-k3.1-pro
• $1.78 per 1M tokens right now
• OpenRouter charges $4.00
• Down 1.0% in the last 24h
Which model should we chart next?
https://qubax.ai/market/kimi-k3.1-pro
you could run a top open weights model like kimi k3 via openrouter (or rent a cluster from runpod etc if you don't want to pay a provider)
November 2, 2026 at 9:27 AM
you could run a top open weights model like kimi k3 via openrouter (or rent a cluster from runpod etc if you don't want to pay a provider)
For this exact workflow you need ~32GB RAM. The more exciting thing is that these models run *very* cheaply on OpenRouter and similar sites (eg: $0.015 in / $0.83 out for Qwen3.8 27b). This is >10x cheaper than Sonnet 5.5.
November 2, 2026 at 2:03 PM
For this exact workflow you need ~32GB RAM. The more exciting thing is that these models run *very* cheaply on OpenRouter and similar sites (eg: $0.015 in / $0.83 out for Qwen3.8 27b). This is >10x cheaper than Sonnet 5.5.
Un bug dans Zed (l'éditeur) m'a coût 2000$ (en réalité 1000$ en septembre et 1000$ en octobre). 💸
Je n'avais même pas fait gaffe : le cache OpenRouter était désactivé. 🥹
Bref, patch en cours. En attendant, si vous utilisez Zed + OpenRouter + Claude : n'utilisez pas les alias ~latest !
Je n'avais même pas fait gaffe : le cache OpenRouter était désactivé. 🥹
Bref, patch en cours. En attendant, si vous utilisez Zed + OpenRouter + Claude : n'utilisez pas les alias ~latest !
November 3, 2026 at 9:44 PM
Un bug dans Zed (l'éditeur) m'a coût 2000$ (en réalité 1000$ en septembre et 1000$ en octobre). 💸
Je n'avais même pas fait gaffe : le cache OpenRouter était désactivé. 🥹
Bref, patch en cours. En attendant, si vous utilisez Zed + OpenRouter + Claude : n'utilisez pas les alias ~latest !
Je n'avais même pas fait gaffe : le cache OpenRouter était désactivé. 🥹
Bref, patch en cours. En attendant, si vous utilisez Zed + OpenRouter + Claude : n'utilisez pas les alias ~latest !