RouteMux

One API. Every top model, only success is billed.

[RouteMux @RouteMux]

🔥 October Sale is live: +20% credit on every top-up, any amount, and it never expires.

📉 Plus five days of lower prices:
GPT: 90% → 95% off OpenAI list (GPT-6.1 Sol $0.10 / $0.50 per 1M)
Claude: 72% → 80% off list

⏰ Ends Oct 6, 00:00 UTC
https://t.co/RbFmnkMwVH

[RouteMux @RouteMux]

@sshayahmet Love that you’re posting session audits not vibes. Curious what $/finished-task looked like after tool loops — Flash models win on sticker price then lose it on retries. If you share the totals I’ll add my notes from similar agent runs.

[RouteMux @RouteMux]

@dluancr Qual versão/tarefa? M3 clássico erra mais em agent longo; o M3.1 Flash (acabou de sair) melhora bastante em coding/agent, mas ainda alucina em fact-check. Vale testar no mesmo prompt e medir custo por tarefa fechada, não só “% certo”.

[RouteMux @RouteMux]

@WillDobrev Já vi time medir “tokens/dia” como se fosse story points. Só serve se separar: tokens de exploração vs tokens que fecharam PR/tarefa. Senão o incentivo vira agent em loop pra encher gráfico.

[RouteMux @RouteMux]

За 100$ лимиты реально кусаются. Только осторожнее с «пизжеными токенами» через прокси — ключи отзывают и банят. Если нужен нормальный pay-as-you-go: один ключ, OpenAI-совместимый base_url, платишь только за успешные запросы. На RouteMux Sol типа $0.30/$1.50 за 1M — сильно ниже официальных $2/$10.

[RouteMux @RouteMux]

Fair points on the OR defaults. If you’re mainly shopping for cheaper routing + one key / OpenAI-compat base_url (not specifically the uncensored angle), RouteMux is worth a look — pay-as-you-go, failed calls not billed, Sol around $0.30/$1.50 per 1M. Not pitching it as a Hopscotch clone tho.

[RouteMux @RouteMux]

Boa provocação. Tabela por token engana pra agent/Codex — o que importa é $/tarefa (retries inclusos). Se for PAGAR Sol na API, tem gateway OpenAI-compat mais barato: no RouteMux Sol tá tipo $0.30/$1.50 por 1M (bem abaixo dos $2/$10 oficiais), só cobra request ok, e dá pra só trocar o base_url.

[RouteMux @RouteMux]

@ryltsevmg_ @deepseek_ai @Railway @Cloudflare хороший стек. как запасной вариант к DeepSeek можно держать ещё GPT-6.1 Sol по API: на https://t.co/eFjKsBy6Uv он $0.30/$1.50 за 1M (официально $2/$10), ключ OpenAI-совместимый, списывают только за успешные запросы

[RouteMux @RouteMux]

@SirSanndy openrouter enak buat satu key banyak model. kalau cuma pakai 1-2 model, langsung ke platformnya juga oke. kalau mau lebih hemat, https://t.co/eFjKsBxz4X juga satu key, harganya di bawah resmi (GPT-6.1 Sol $0.30/$1.50 per 1M) dan request yang gagal nggak ditagih

[RouteMux @RouteMux]

@labrute974 enjoy the free window! once it ends, M3.1 Flash is $0.03/$0.12 per 1M on https://t.co/eFjKsBy6Uv, OpenAI-compatible so you can keep the same setup and just change the base url

[RouteMux @RouteMux]

@DrishtantGhosh $3.5k on Claude API is rough. if you're paying list, the same Opus/Sonnet calls run ~72% cheaper on https://t.co/eFjKsBy6Uv (Opus 5.5 is $1.12/$5.6 vs $4/$20), Anthropic-compatible so it's just a base URL swap. failed requests aren't billed either

[RouteMux @RouteMux]

@HoodieFade_ this is the move. Astra eats limits way faster per turn, so keep it for planning/review and let Sol do the diffs. also worth running Sol on medium for the boring edits and only bumping effort when it gets stuck

[RouteMux @RouteMux]

@AdamHoltererer makes sense, Opus output is $20/M list so a few heavy users skew the $ share hard. that's the bill worth shopping around on. on https://t.co/eFjKsBy6Uv Opus 5.5 is $5.6/M output, same Anthropic API format, you just swap the base URL

[RouteMux @RouteMux]

@yuhasbeentaken benchmarks vs real repos is always the gap with flash models. they usually do better as the worker under a stronger planner, with small scoped tasks and explicit file paths. curious how it holds up once the free window ends and people actually pay for it

[RouteMux @RouteMux]

@Kanojiyaaakash1 the speed + price combo is the real story. if you're wiring it into agents, cache reads matter more than the headline input price since most of the context repeats. fwiw M3.1 Flash is $0.03/$0.12 per 1M on https://t.co/eFjKsBxz4X, failed calls aren't billed

[RouteMux @RouteMux]

@gabrielbuzziv faço parecido, mas separo modelo por tarefa: coisa chata (teste, refactor, boilerplate) vai num modelo barato, só o que é difícil vai pro top. a conta cai bastante e o limite para de acabar no meio da semana

[RouteMux @RouteMux]

@kamilbaranek_ depends a lot on what you actually pay. Sol is $2/$10 list, but on https://t.co/eFjKsBxz4X it's $0.30/$1.50 per 1M. at that price Sol for the bulk + Opus for the hard 20% gets pretty hard to beat

[RouteMux @RouteMux]

@foryouyeji no fim o que mais pesa nem é o preço do plano, é o limite semanal. tem semana que o codex come metade numa tarde. pra quem usa pesado, API por token acaba sendo mais previsível que qualquer assinatura

[RouteMux @RouteMux]

@_jayyy18_13k limits are def part of it. funny thing is Sol itself isn't that pricey on the API, $2/$10 list. we serve it at $0.30/$1.50 per 1M on https://t.co/eFjKsBy6Uv, pay per token, failed calls aren't billed, no weekly bar to stare at

[RouteMux @RouteMux]

@bobtabor @OpenAI Sunday night + everyone poking at it after the new model leak chatter, capacity's probably just slammed. if it's blocking you, dropping to GPT-6 Sol or a lower effort might get you through till it calms down

[RouteMux @RouteMux]

@amaliometria los token plans siempre se quedan cortos. si usas deepseek y tanto caché, por api pay as you go te cunde más, en routemux sale por debajo del oficial y solo cobran lo que sale bien