Production AI engineering guides.
Long-form playbooks on the topics that actually decide whether your AI feature ships profitably. Cost, architecture, compliance, multimodal — each guide is the canonical Kunavo answer plus links to every supporting deep dive.
- 문제 해결·8 min read
클로드 코드 오류 정리 — 401·429·529와 설치 오류 구분법
클로드 코드 오류는 설치·실행 단계의 클라이언트 오류와, 모델을 호출할 때 돌아오는 API 오류로 나뉩니다. 원인도 해결 방법도 다르므로, 먼저 어느 쪽인지 가르는 것이 가장 빠릅니다.
Read guide - 설정·7 min read
클로드 API 키 발급 방법 — 공식 경로와 게이트웨이 경로
클로드 API 키를 받는 경로는 두 가지입니다. 둘 다 같은 Claude 모델을 호출하고, 코드에서 달라지는 것은 base URL과 키 문자열뿐입니다.
Read guide - Setup·8 min read
Cline vs Claude Code — the billing shapes, not the vibes
Both send requests to an endpoint you choose. What differs is how each is paid for, and how much context each carries per step.
Read guide - Pricing·8 min read
What a Claude Code workflow costs — the fan-out nobody prices
A workflow's cost is its fan-out. That is the point of it, and it is also the bill.
Read guide - Pricing·8 min read
Codex vs ChatGPT — two doors, one allowance, very different burn
The interfaces are well covered elsewhere. This is what the choice costs.
Read guide - Pricing·9 min read
Codex pricing — what a task costs per token, and the break-even against a plan
Plan or per-token is one division, not a debate. Here are the rates, one worked task, and the break-even.
Read guide - Pricing·10 min read
Claude Pro and Max limits — what happens when you hit one, and what it costs
Every guide explains the limits. This one answers what to do at four in the afternoon when you hit one.
Read guide - Models·9 min read
Claude Opus vs Sonnet vs Haiku — cost per finished task, not cost per token
A tier that costs 5x per token but finishes in one attempt instead of three is not 5x the cost. Here is the arithmetic.
Read guide - Setup·9 min read
--dangerously-skip-permissions — the two blast radii, and three containments
An agent that never stops to ask is also an agent that never stops spending. Contain both, then use the flag.
Read guide - Pricing·10 min read
Claude prompt caching — the rates, the usage object, and why hit rates drop behind a gateway
Marked correctly it is a tenth of the rate. Marked wrong it costs more than not caching at all.
Read guide - Pricing·7 min read
Nano Banana Pro pricing — per-image rates and what actually moves the bill
Per image, not per token — and resolution is the lever most people leave on a default.
Read guide - Setup·8 min read
LibreChat vs Open WebUI — which takes a custom endpoint with less friction
Two environment variables versus a YAML block — and what the YAML buys you.
Read guide - 費用·9 min read
Claude Code 費用 2026 — 訂閱與 API 按量計費、實際金額與損益平衡點
Claude Code 的錢有兩種算法 —— 固定月費的訂閱，或按 token 計費的 API 金鑰 —— 而「到底多少錢」會吵不完，多半就是這兩種被混在一起講。這頁把兩邊現在的數字、一個月的實際金額、以及怎麼選的算式放在同一張表上。
Read guide - 요금·9 min read
Claude Code 요금 2026 — 구독 vs API 종량제, 실제 금액과 손익분기점
Claude Code 요금이 매번 헷갈리는 이유는 고정 월정액 구독과 토큰 종량제 API 키가 섞여서 이야기되기 때문입니다. 양쪽의 현재 숫자, 한 달 실제 금액, 어느 쪽을 고를지 계산하는 법을 한 장에 정리했습니다.
Read guide - 가격·8 min read
GPT API 가격 2026 — GPT-5.4 / 5.5 / 5.6 Sol 토큰 요율과 실제 비용
GPT API는 무료 등급 없이 첫 호출부터 토큰 단위로 과금됩니다. GPT-5 계열의 현재 요율을 OpenAI 정가와 나란히 놓고, 청구액을 실제로 좌우하는 두 가지 함정과 한국에서 결제하는 방법을 정리했습니다.
Read guide - 料金·9 min read
Claude Code 料金【2026年版】— サブスクと API 従量課金、実額と損益分岐点
Claude Code の料金は「定額の月額」か「トークン従量課金」のどちらかで、値段の話がこじれる原因はほぼこの 2 つの混同にあります。両方の現在の数字、1 か月の実額、どちらを選ぶかの計算をまとめました。
Read guide - Prix·9 min read
Prix de Claude Code 2026 — abonnement ou API, coût réel et point d'équilibre
Claude Code coûte soit un forfait mensuel, soit une facture au token, et presque toute réponse confuse sur son prix vient du mélange des deux. Voici les chiffres actuels des deux voies, ce que coûte un vrai mois, et le calcul pour choisir.
Read guide - Pagamento·6 min read
Como pagar a API do Claude com Pix — passo a passo, sem cartão internacional
No Brasil o obstáculo para usar a API do Claude quase nunca é o código — é o cartão internacional. Este é o caminho por Pix, do cadastro à primeira resposta do modelo, com o valor em reais e compensação na hora.
Read guide - Troubleshooting·6 min read
「メッセージストリームでエラーが発生しました」の原因と対処法 — ChatGPT・AI チャット共通
「メッセージストリームでエラーが発生しました」は、ChatGPT がサーバーから回答を少しずつ受け取るストリーミング通信が途中で切断されたときに表示されるエラーです。エラーの原因はユーザーの操作ミスではなく、サーバー側の過負荷・通信の不安定・ブラウザ環境の問題がほとんどを占めます。多くのケースは回答の再生成かページの再読み込みで解消します。解消しない場合は、障害情報の確認 → 回線とブラウザの切り分け → 新しいチャットへの移行、の順に試すのが最短です。
Read guide - Troubleshooting·6 min read
"API key not valid. Please pass a valid API key." — the five things Gemini means by it
This message is Google's catch-all: the key you sent could not be used for this call. That is not the same as the key being wrong, and four of the five causes leave the key itself perfectly valid — which is why re-copying it is usually wasted effort.
Read guide - Troubleshooting·6 min read
Anthropic API 404 not_found_error: "model: …" — why a model name that exists still 404s
The confusing part of this 404 is that the model usually does exist — in Anthropic's docs, in a blog post, in last quarter's code. What it does not exist in is the list your API key is allowed to address today.
Read guide - Troubleshooting·6 min read
ANTHROPIC_AUTH_TOKEN vs ANTHROPIC_API_KEY — which one Claude Code actually reads
These are not aliases. One becomes `x-api-key`, the other becomes `Authorization: Bearer` — and when both are set, the one you did not mean to use can win. Most "my key is correct but auth fails" reports against a custom base URL are this.
Read guide - Troubleshooting·6 min read
Claude Code "API Error: bad_response_status_code" — reading the status underneath it
This error tells you the call failed and almost nothing about why. The useful information — the status code and the provider's message — is one debug flag away, and each status points at a different fix.
Read guide - Troubleshooting·6 min read
Erro 529 overloaded_error na API da Claude — o que significa e como contornar
O 529 é o único erro da Claude que o seu código não causou: a própria Anthropic está sobrecarregada. Não dá para corrigir — dá para absorver bem. Isso significa retry paciente, um modelo de reserva e nunca amplificar o incidente com tentativas imediatas.
Read guide - Troubleshooting·6 min read
Erro 429 rate_limit_error na API da Claude — o que significa e como resolver
429 é o erro que diz "mais devagar", não "você está sem dinheiro". A diferença importa: a correção de um é esperar, a do outro é recarregar — e confundir os dois leva a horas de depuração no lugar errado.
Read guide - Troubleshooting·6 min read
Erro 401 authentication_error / invalid x-api-key — o que verificar, na ordem
Quase todo 401 tem uma de quatro causas, e só uma delas é "a chave está errada". As outras três deixam a chave perfeitamente válida — por isso recriar a chave costuma ser esforço desperdiçado.
Read guide - Troubleshooting·6 min read
Claude API 500 api_error and 502 Bad Gateway — a retry policy that actually recovers
500 and 502 mean a fault, not a limit. That makes them the one error class in this family worth retrying almost immediately — unlike 429, which is your own rate limit, and 529, which is the provider being full. Retrying the wrong one of the three turns a small incident into your incident.
Read guide - Troubleshooting·6 min read
Claude Code “API Error: 401 authentication_error” with a custom base URL — every cause
A 401 here is about your credential or your base URL, never your model — a model problem returns 404 with a message naming the model. That single distinction resolves most of these in one curl.
Read guide - Troubleshooting·6 min read
Claude API 400 “tool_use ids were found without tool_result blocks” — the ordering rule
This is a message-ordering error, not a tools error. Claude requires every tool_use block in an assistant turn to be answered by a tool_result block in the immediately following user turn — same ids, nothing in between. Your loop dropped one, usually because the tool threw.
Read guide - Troubleshooting·6 min read
“Unsupported parameter: 'max_tokens' is not supported with this model” — use max_completion_tokens
The rename is the easy half. The half that catches people is what the new field counts — max_completion_tokens covers reasoning plus visible output, so a budget sized for the answer alone comes back empty with finish_reason "length", and billed.
Read guide - Troubleshooting·6 min read
Claude API request timeouts and “streaming is strongly recommended” — the 10-minute rule
There are usually three timeouts between you and the model — your SDK's, an intermediary's, and the provider's own rule about long non-streaming calls — and the shortest one wins. Raising only the SDK's is why this keeps happening after you thought you fixed it.
Read guide - Troubleshooting·6 min read
Gemini API 400 INVALID_ARGUMENT — “please use a valid role” and the empty-contents family
Nearly every INVALID_ARGUMENT of this shape comes from posting OpenAI-style messages to Google's native generateContent endpoint. It has different role names, a separate home for system instructions, and no tolerance for an empty turn.
Read guide - Troubleshooting·6 min read
OpenAI “You exceeded your current quota” (insufficient_quota) — why waiting never fixes it
This is the 429 that backoff cannot fix. insufficient_quota means your account has no spendable credit — the request was rejected before any model ran, and it will keep being rejected until billing changes. Here is how to confirm which billing state you're in and clear it in minutes.
Read guide - Troubleshooting·6 min read
Gemini 503 “The model is overloaded” (UNAVAILABLE) — retries that work and when to stop
“The model is overloaded” is the one Gemini error that isn't about you: the serving pool for that model has no capacity right now. Nothing in your project settings can prevent it — what you control is how gracefully you ride it out, and which fallback you reach for when it doesn't pass.
Read guide - Preise·14 min read
Claude Preise 2026 — was kostet Claude? Abo, API und Claude Code
Der Chat ist im Free-Plan kostenlos, Claude Pro kostet $20 im Monat und Claude Max beginnt bei $100 — über die API zahlen Sie statt einer Pauschale pro Token, auf Kunavo 30–60% unter Anthropics Liste. Diese Seite rechnet Abo, API und Claude Code nebeneinander durch.
Read guide - Integração·8 min read
Kilo Code com a API do Claude — base URL, modelos e custo (também Cline e Roo Code)
Kilo Code, Cline e Roo Code aceitam um endpoint compatível com OpenAI personalizado: três campos bastam para rodar Claude dentro do VS Code, cerca de 60% abaixo da tabela Anthropic. O passo a passo por ferramenta, qual modelo escolher, o custo real de uma sessão e como pagar do Brasil por Pix.
Read guide - Preços·9 min read
Claude Code preço 2026 — assinatura vs API, tarifas por modelo e pagamento com Pix
O Claude Code custa ou uma mensalidade fixa ou uma conta por token, e quase toda resposta confusa sobre o preço dele vem de misturar as duas. Aqui estão os números atuais das duas rotas, quanto custa um mês de verdade, a conta para escolher — e como pagar do Brasil sem cartão internacional.
Read guide - Integração·8 min read
Codex CLI com chave de API — configuração, modelos e custo de sessão
O Codex CLI roda com login do ChatGPT ou com chave de API — e a rota de chave é a única que permite apontá-lo para outro provedor ou outra família de modelos. A configuração que funciona, o requisito que trava a maioria dos gateways, o custo de uma sessão e como pagar por Pix.
Read guide - Integration·9 min read
Cline with the Claude API — base URL setup, Plan/Act model routing, real costs
Cline brings its own agent loop but no model, so it needs an API key. Here is the three-field setup, the base-URL difference between its two provider routes, and how Plan/Act mode turns model choice into a cost lever.
Read guide - Integration·9 min read
Roo Code with the Claude API — provider setup and a model per mode
Roo Code ships modes but no model. Here is the three-field provider setup, the base-URL trap on its native Anthropic route, and how configuration profiles let Architect think on Opus while Code executes on Sonnet.
Read guide - Pricing·9 min read
Claude Code pricing 2026 — subscription vs API, per-model rates, and what a month costs
Claude Code costs either a flat monthly fee or a per-token bill, and almost every confusing answer about its price comes from mixing the two up. Here are the current numbers for both routes, what a real month costs, and the arithmetic for choosing.
Read guide - Setup·8 min read
Install Claude Code — the command for every OS, first login, and the errors people hit
Installing Claude Code is one command on every platform. Here is that command for each OS, what to run when it doesn't work, and the four environment variables that put it on a different endpoint afterwards.
Read guide - Troubleshooting·9 min read
OpenAI API rate limits — which limit you hit, how to read it, and the retry that fixes it
A 429 from OpenAI means one of five ceilings was crossed, and the fixes point in opposite directions depending on which. Here is how to read the answer straight off the response headers, and the retry logic that actually stops the errors.
Read guide - Integration·9 min read
Claude Code Router — route Claude Code to any model, or skip the router entirely
Most people reaching for claude-code-router only want Claude Code somewhere cheaper — and that is a base-URL swap, not an install. Here is when the router actually earns its place, the exact variables for both routes, and an honest list of what a gateway breaks.
Read guide - Integration·8 min read
Claude Code API key — where to get one, where to put it, and why the wrong variable fails silently
Claude Code takes a subscription login or an API key, and the two behave very differently once set. Here is where each comes from, exactly which variable holds it, and the header mismatch behind most 401s.
Read guide - Cost·7 min read
Is Claude Code free? What's free, what isn't, and what a month costs
Claude Code is free to install and not free to run. Here is exactly where the line sits, what a real month of agentic coding costs, and the cheapest legitimate way to run it.
Read guide - Comparison·8 min read
Claude Code vs Codex CLI — what actually differs, and how to choose
Claude Code and Codex CLI do the same job from the same place. The durable differences are model family, extension surface and how freely you can point each at your own endpoint — that last one is sharper than most comparisons admit.
Read guide - Setup·9 min read
Codex CLI API key — the working custom-provider setup
Codex CLI accepts a custom provider only over the Responses API. Here is the config block that works, what each field does, and what a session costs.
Read guide - Architecture·13 min read
Invoice Data Extraction: deep learning vs machine learning vs LLM (2026 guide)
Invoice extraction has run through four generations of technique — templates, classical machine learning, fine-tuned deep learning, and now a single vision-LLM call. Here is how they compare, and the complete working LLM pattern: code, structured output, guardrails and per-invoice economics.
Read guide - Preços·8 min read
Claude API preço 2026 — tarifas por modelo, pagamento com Pix e custos reais
Claude é o padrão para raciocínio, código e agentes. Aqui estão as tarifas atuais da API do Claude por modelo — 30–60% abaixo da tabela Anthropic — com pagamento do Brasil (Pix, cartão internacional, Apple Pay), exemplos de custo para verificar por conta própria e as alavancas que mais reduzem a fatura.
Read guide - Fiyatlandırma·8 min read
Claude API fiyatları 2026 — model bazında tarifeler, ödeme ve gerçek maliyetler
Claude; akıl yürütme, kod ve ajanların standardı. İşte model bazında güncel Claude API tarifeleri — Anthropic fiyatlarının %30–60 altında — Türkiye'den ödeme (kart, Apple Pay), kendinizin doğrulayabileceği maliyet örnekleri ve faturayı en çok düşüren yöntemlerle birlikte.
Read guide - Harga·8 min read
Claude API harga 2026 — tarif per model, cara bayar, dan biaya nyata
Claude adalah standar untuk penalaran, kode, dan agen. Inilah tarif terbaru API Claude per model — 30–60% di bawah harga resmi Anthropic — beserta cara bayar dari Indonesia (kartu internasional, Apple Pay), contoh biaya yang bisa Anda verifikasi sendiri, dan cara paling efektif menekan tagihan.
Read guide - Troubleshooting·6 min read
Claude API 429 rate_limit_error — causes and the fix that holds
A 429 from the Claude API means you crossed one of Anthropic's per-minute limits — requests, input tokens, or output tokens. The fix is rarely "wait longer": it's honoring retry-after, adding jittered backoff, and smoothing bursts. Here is the complete playbook.
Read guide - Troubleshooting·6 min read
Claude API 401 authentication_error / invalid x-api-key — every cause
A 401 from Claude is always one of five things: the wrong header, the wrong key type for the endpoint, a malformed env var, a revoked key, or the wrong base URL for that key. Run the diagnostic below and you'll find yours in under a minute.
Read guide - Troubleshooting·6 min read
Claude API “credit balance is too low” / 402 insufficient_quota — the fix
This error has nothing to do with your code: the prepaid balance behind the key is empty (or your monthly spend cap tripped). Here's how the billing models differ between Anthropic Console and gateways, and the alarms that stop it recurring at 2 a.m.
Read guide - Troubleshooting·6 min read
Claude API 529 overloaded_error — what it is and how to ride it out
529 is the one Claude error your code didn't cause: Anthropic itself is overloaded. You can't fix it — you can only absorb it gracefully. That means patient retries, a fallback model, and never amplifying the incident with instant-retry storms.
Read guide - Troubleshooting·6 min read
Gemini API key not working — API_KEY_INVALID and its five causes
"API key not valid. Please pass a valid API key." — Gemini's least helpful sentence. The key usually IS valid; something around it is wrong: a referrer restriction, the Generative Language API not enabled, or an AI-Studio key sent to a Vertex endpoint. Work the list below in order.
Read guide - Troubleshooting·6 min read
Gemini API 429 RESOURCE_EXHAUSTED — quota vs rate limit, fixed properly
RESOURCE_EXHAUSTED is Gemini's catch-all 429. If it clears after a minute, you hit a rate limit — backoff fixes it. If it lasts until midnight Pacific, you exhausted a daily free-tier quota — no retry loop can fix that; only billing or a different route can.
Read guide - Troubleshooting·6 min read
OpenAI-compatible API returning 401/403 — base_url and header pitfalls
The whole point of OpenAI-compatible APIs is that the SDK just works — so when it 401s, the bug is almost always in the two lines you changed: base_url and api_key. Here are the failure modes in the order they actually occur.
Read guide - Troubleshooting·6 min read
context_length_exceeded / prompt is too long — fixes that don't lobotomize your app
Every model has a context window; this 400 means prompt + requested output didn't fit. The naive fix — truncating the prompt blindly — is how RAG apps silently get dumber. Here's how to fit the window without losing the plot.
Read guide - Troubleshooting·6 min read
LLM streaming errors — SSE cutoffs, hanging streams and missing usage
Streaming failures are rarely the model — they're the plumbing between you and it. Idle-timeout proxies kill quiet connections, nginx buffers eat events, and half-consumed streams look like "the API stopped responding". Work through the plumbing first.
Read guide - Troubleshooting·6 min read
model_not_found / 404 — model naming across Claude, Gemini and gateways
Model ids are not portable: api.anthropic.com wants dated ids, Gemini wants its own version strings, and every gateway defines slugs. A 404 here means "this host has no model by that exact string" — the fix is always to ask the endpoint what it serves.
Read guide - Prix·8 min read
Prix de l'API Claude 2026 — tarifs par modèle, paiement et coûts réels
Claude est le standard pour le raisonnement, le code et les agents. Voici les tarifs actuels de l'API Claude par modèle — 30 à 60 % sous les prix Anthropic — avec le paiement depuis la France (CB, TVA), des exemples de coût à vérifier soi-même et les leviers qui réduisent le plus la facture.
Read guide - Prix·8 min read
Prix de l'API Gemini 2026 — tarifs par modèle, exemples et accès moins cher
Gemini offre l'un des meilleurs rapports qualité-prix de l'IA de pointe. Voici les tarifs actuels par modèle — environ 70 % sous le prix catalogue de Google — avec des exemples de coût et la façon la moins chère d'appeler Gemini en production.
Read guide - Clé API·7 min read
Clé API Gemini — comment l'obtenir et l'utiliser avec le SDK OpenAI
Deux façons de créer une clé API Gemini — Google AI Studio (officielle) ou une seule clé Kunavo compatible OpenAI, sans projet Google Cloud — et le premier appel en quelques minutes.
Read guide - Intégration·8 min read
Kilo Code avec l'API Claude — configuration base URL (aussi Cline et Roo Code)
Kilo Code, Cline et Roo Code acceptent tous un endpoint compatible OpenAI personnalisé : trois champs suffisent pour faire tourner Claude dans VS Code, environ 60 % sous le tarif Anthropic. Le pas-à-pas par outil, le choix du modèle et le coût réel d'une session.
Read guide - Integration·8 min read
Kilo Code with the Claude API — OpenAI-compatible base URL setup and per-mode routing
Kilo Code accepts a custom OpenAI-compatible endpoint — three settings fields put Claude in your editor at about 60% under Anthropic's list price. Here's the setup, which model to pick per mode, and what an agentic coding session actually costs.
Read guide - Architecture·8 min read
LLM gateway — one API for every model: routing, fallback, billing & observability (2026)
An LLM gateway is one API in front of every model provider — routing, failover, caching, one bill and per-key spend control. This is what it does, how it differs from an API gateway, the two architectures, and how Kunavo implements it OpenAI-compatibly.
Read guide - Video·7 min read
Text-to-video API — generate video from a prompt (OpenAI-compatible, 2026)
A text-to-video API turns a prompt into a video clip programmatically. Here's the API shape, the live models on Kunavo (Veo 3.1, Seedance 2, Wan 2.7), per-clip pricing, and the async pattern you need in production.
Read guide - Video·7 min read
Veo 3.1 API — Google's text-to-video with native audio, OpenAI-compatible (2026)
Veo 3.1 is Google's text-to-video model with native audio. Here's how to call the Veo 3.1 API on Kunavo — three tiers, image-to-video, per-video pricing ~50–70% under Google's list, and the async workflow.
Read guide - Image·6 min read
GPT-Image-2 API — OpenAI's image model at half the price, OpenAI-compatible
GPT-Image-2 is OpenAI's latest image model. Here's how to call the GPT image API through an OpenAI-compatible endpoint — text-to-image, image edit, per-image pricing at about half of list, on one key.
Read guide - Image·6 min read
Seedream API (incl. Seedream 4.5) — what it is, pricing, and the live alternative
Seedream is ByteDance's text-to-image family, known for character consistency. Here's what the Seedream and Seedream 4.5 APIs do, how they're priced, and how to run the same workflow on Kunavo today with Nano Banana or GPT-Image-2.
Read guide - Video·6 min read
Kling API & Kling AI API pricing — what it is, and the live alternative on Kunavo
Kling is Kuaishou's video model family, known for physical realism. Here's what the Kling API does, how its credit-based pricing works, and how to run text-to-video on Kunavo today with Google Veo 3.1.
Read guide - Audio·6 min read
Suno API documentation — generate music from a prompt (V5 / V5.5), with pricing
Suno generates full songs — vocals, lyrics and instrumentation — from a prompt. Here's the Suno API documentation: the /v1/audio/music endpoint, custom lyrics, per-request pricing, and the async workflow.
Read guide - Compare·9 min read
OpenRouter alternatives (2026) — an honest roundup: cheaper, multimodal, BYO-key and open-source options
OpenRouter is the biggest text-LLM router, and sometimes still the right call. This is the honest map of alternatives — by motivation: lower prices, multimodal endpoints, your own keys under a control plane, or open source you can self-host.
Read guide - Models·6 min read
Gemini 3 API — pricing, availability, and the live way to call Gemini today
Gemini 3 is Google's flagship family — 3 Pro, 3.1 Pro and 3 Flash. Here's its API pricing at Google's list, its honest availability status on Kunavo, and the live Gemini 2.5 route that upgrades to Gemini 3 with a one-word change.
Read guide - Video·7 min read
Seedance 2 API — pricing per second, code, and every model (2026)
Seedance 2 is ByteDance's current text-to-video family — native audio, multi-shot storytelling, reference consistency. Four variants are live here on one OpenAI-style endpoint, billed per second in USD.
Read guide - Video·6 min read
Wan API — Alibaba's Wan 2.7 text-to-video, live with per-second pricing
Wan is Alibaba's cinematic text-to-video family; Wan 2.7 is the current generation and it's live on Kunavo. Per-second USD pricing, one OpenAI-compatible key, and a one-word migration from any earlier Wan integration.
Read guide - Image·5 min read
Grok Imagine API — what it is, xAI's pricing, and the live alternatives
Grok Imagine is xAI's image and short-video generator with a distinctive aesthetic. Here's what its API costs at xAI's list, its honest availability status on Kunavo, and the live image models that cover the same workloads today.
Read guide - Video·6 min read
Hailuo API (MiniMax video) — models, credit pricing, and the live alternative
Hailuo is MiniMax's video model family, known for smooth, physically plausible motion. Here's what the Hailuo API does, how its credit pricing works, and how to run the same workloads on Kunavo today with Veo 3.1.
Read guide - Image·5 min read
FLUX.2 API — Black Forest Labs' pricing, and the live image models today
FLUX.2 is Black Forest Labs' second-generation image family, priced aggressively at $0.03 per image. Here's what its API does, its honest availability status on Kunavo, and the live models covering the same workloads today.
Read guide - 動画·7 min read
Seedance API の使い方 — 料金と Python 実装(Seedance 2 対応・2026)
ByteDance の動画生成モデル Seedance 2 を、OpenAI 互換 API でそのまま呼び出す方法。料金表・Python コード・Fast との使い分け・Veo 3.1 との比較まで、本番導入に必要な情報をまとめました。
Read guide - 영상·7 min read
Seedance API 가이드 — 요금과 Python 사용법 (Seedance 2 지원, 2026)
ByteDance의 영상 생성 모델 Seedance 2를 OpenAI 호환 API로 바로 호출하는 방법. 요금표, Python 코드, Fast와의 사용 구분, Veo 3.1 비교까지 프로덕션 도입에 필요한 내용을 담았습니다.
Read guide - Pricing·8 min read
OpenAI GPT API pricing 2026 — GPT-5.4, 5.5, Mini & Codex costs, examples, cheaper access
GPT-5 is OpenAI's frontier family. Here are the current GPT API prices per model — about 60% below OpenAI's list — with worked cost examples, the official OpenAI reference, and the cheapest way to call GPT in production.
Read guide - Models·8 min read
Claude vs GPT vs Gemini (2026) — which AI model to use, and what each costs
There's no single best model — match it to the job. Here's the honest positioning of Claude, GPT and Gemini, a like-for-like price table, a which-to-use guide, and how to test all three from one key.
Read guide - Image·6 min read
Nano Banana API — call Google's image models OpenAI-compatibly at half the price
Nano Banana is Google's image model family. Here's how to call it through an OpenAI-compatible API — text-to-image, image edit, per-model pricing at roughly half of list, on one key.
Read guide - API·7 min read
OpenAI-compatible API guide — from Ollama's local endpoint to hosted frontier models
'OpenAI-compatible' means any OpenAI SDK works by changing only base_url. Ollama exposes one for local models; Kunavo exposes one for hosted frontier models. Here's the pattern, and how to move between them without rewriting code.
Read guide - Architecture·8 min read
What is an AI gateway? The LLM gateway pattern, explained (2026)
An AI gateway is one API in front of every model provider — so your app calls a single endpoint and key instead of wiring Anthropic, Google and OpenAI separately. This is what a gateway does, when you actually need one, and how Kunavo implements the pattern.
Read guide - 가격·8 min read
Claude API 가격·결제 2026 — 클로드 API 비용 완전 정리
한국 개발자가 가장 많이 찾는 '클로드 API 결제·비용·가격'을 한곳에 정리했습니다. Kunavo는 Claude 전 모델을 Anthropic 정가보다 약 60% 저렴하게 제공하고, 국내 카드·Apple Pay로 최소 $10부터 충전해 종량제로 결제합니다 — 잔액은 만료되지 않습니다.
Read guide - Preços·8 min read
Preços da API Gemini 2026 — custos por modelo, exemplos e acesso compatível com OpenAI mais barato
O Gemini está entre as melhores relações custo-benefício da IA de ponta. Veja as tarifas atuais por modelo — cerca de 70% abaixo da lista do Google — com exemplos de custo e a forma mais barata de chamar o Gemini em produção.
Read guide - Chave de API·7 min read
Chave de API Gemini — como obter e usar com o SDK da OpenAI
Como criar uma chave de API Gemini pelos dois caminhos — Google AI Studio (oficial) ou uma única chave Kunavo, compatível com OpenAI, sem projeto no Google Cloud — e fazer a primeira chamada.
Read guide - Vídeo·7 min read
API Sora — texto-para-vídeo em um fluxo compatível com OpenAI
O Sora é o modelo de texto-para-vídeo da OpenAI. Na Kunavo, o modelo ao vivo é o Google Veo 3.1, no mesmo endpoint de vídeo no estilo OpenAI — eis o fluxo de trabalho, com o acesso ao Sora no roadmap.
Read guide - Preise·8 min read
Gemini API Preise 2026 — Tarife pro Modell, rund 70% unter Googles Liste
Aktuelle Gemini API Preise pro Modell, durchgerechnete Kostenbeispiele zum Selbstprüfen und der günstigste Weg, Gemini in Produktion aufzurufen — rund 70% unter Googles Liste hinter einer OpenAI-kompatiblen API.
Read guide - API-Key·8 min read
Google Gemini API Key erstellen — zwei Wege (Google AI Studio & Kunavo)
So erstellen Sie einen Google Gemini API Key — der offizielle Weg über Google AI Studio und der Ein-Key-Weg über Kunavo, der ohne Google-Cloud-Projekt zugleich Claude, GPT sowie Bild und Video erreicht.
Read guide - Video·7 min read
Sora API — KI-Video generieren am OpenAI-artigen Endpoint (heute mit Veo 3.1)
KI-Video über die Sora API: Sora ist noch nicht aktiviert — das aktive Text-zu-Video-Modell ist Google Veo 3.1 am selben OpenAI-artigen Endpoint. Mit lauffähigen Beispielen und Preisen ab $0.18 pro Clip.
Read guide - Fiyatlandırma·8 min read
Gemini API fiyatlandırma 2026 — model bazlı fiyatlar ve maliyeti düşürme
Gemini fiyatlandırmasının net tablosu: model bazlı oranlar, gerçek dünya maliyet örnekleri ve Gemini'yi üretimde token başına en az fiyata çağırmanın yolu. Tüm rakamlar Kunavo'nun faturalandırdığı oranlardır.
Read guide - API Anahtarı·7 min read
Gemini API anahtarı alma — Google AI Studio ve Kunavo ile adım adım
Gemini API anahtarı almanın iki yolu: doğrudan Google AI Studio (resmi) veya tüm modeller için tek bir OpenAI-uyumlu Kunavo anahtarı. Anahtar oluşturma, ilk çağrı ve anahtarı güvende tutma adımlarıyla.
Read guide - Video·7 min read
Sora API — yapay zeka ile video üretimi (bugün Veo 3.1 ile)
Yapay zeka ile video üretmenin pratik yolu: Sora API'nin biçimi, metinden ve görselden videoya iş akışları ve fiyatlandırma. Her örnek bugün Google Veo 3.1 ile çalışır; Sora geldiğinde değişen tek şey model alanı.
Read guide - 가격·8 min read
Gemini API 가격 2026 — 제미나이 API 요금 완전 정리
Gemini는 프런티어 AI 중 가성비 최고 수준이고, Kunavo는 이를 Google 정가보다 약 60% 저렴하게 제공합니다. 모델별 현재 요율과 직접 검증할 수 있는 비용 예시, 프로덕션에서 가장 저렴하게 호출하는 방법을 한곳에 정리했습니다.
Read guide - API 키·7 min read
Gemini API 키 발급 방법 — Google AI Studio vs Kunavo
Gemini API 키를 얻는 길은 두 가지입니다 — Google AI Studio에서 공식 키를 받거나, Kunavo에서 모든 모델을 하나의 OpenAI 호환 키로 호출하는 방법. 두 경로를 비교하고 프로덕션에서 바로 호출을 시작하는 가장 빠른 길을 보여줍니다.
Read guide - 비디오·7 min read
Sora API 가이드 — 지금 바로 쓰는 AI 영상 생성 (Veo 3.1)
Sora API는 프로그래밍 방식으로 영상을 생성하는 방법입니다. Sora 지원은 로드맵에 있고, 현재 Kunavo의 text-to-video는 같은 OpenAI 스타일 엔드포인트로 Google Veo 3.1에서 실행됩니다 — 모든 예제가 지금 바로 동작합니다.
Read guide - 料金·8 min read
Gemini API 料金【2026年版】— モデル別料率と試算例、最安の呼び出し方
Gemini の API 料金をモデル別の実数で。Kunavo は Google 定価より安く、最低 $10 のチャージから従量課金で呼び出せます。料率・試算例・コスト削減のレバーをすべて runnable なコード付きで掲載。
Read guide - APIキー·7 min read
Gemini API キーの取得方法 — Google AI Studio と Kunavo の 2 つの方法
Gemini API キーを取得する 2 つの方法——Google AI Studio（公式）と、Google Cloud 不要で全モデルを 1 キーにまとめる Kunavo。手順、OpenAI SDK でのクイックスタート、キーの安全な管理まで。
Read guide - 動画·7 min read
Sora API の使い方 — 今すぐ動く AI 動画生成（Veo 3.1 / OpenAI 互換）
Sora API での AI 動画生成。Sora はまだ有効化されていないため、今は同じ OpenAI スタイルのエンドポイントで Veo 3.1 が動きます。後で Sora へ切り替えるのは model フィールドの一語変更だけ。すべての例が今すぐ動きます。
Read guide - Harga·8 min read
Harga Gemini API 2026 — biaya per model, contoh, dan akses kompatibel-OpenAI yang lebih murah
Gemini termasuk yang paling bernilai di AI papan atas. Berikut tarif terkini per model — sekitar 70% di bawah list Google — lengkap dengan contoh biaya dan cara termurah memanggil Gemini di produksi.
Read guide - API Key·7 min read
Gemini API key — cara mendapatkan dan memakainya dengan SDK OpenAI
Cara membuat Gemini API key lewat dua jalur — Google AI Studio (resmi) atau satu kunci Kunavo yang kompatibel-OpenAI, tanpa proyek Google Cloud — dan membuat panggilan pertama.
Read guide - Giá·8 min read
Giá Gemini API 2026 — chi phí theo mô hình, ví dụ, và truy cập tương thích OpenAI rẻ hơn
Gemini nằm trong nhóm đáng giá nhất của AI hàng đầu. Đây là mức giá hiện tại theo mô hình — thấp hơn khoảng 70% so với giá niêm yết Google — cùng ví dụ chi phí và cách rẻ nhất để gọi Gemini trong production.
Read guide - API Key·7 min read
Gemini API key — cách lấy và dùng với SDK OpenAI
Cách tạo Gemini API key theo cả hai con đường — Google AI Studio (chính thức) hoặc một khoá Kunavo tương thích OpenAI, không cần dự án Google Cloud — và thực hiện lệnh gọi đầu tiên.
Read guide - Pricing·8 min read
Google Gemini API pricing September 2026 — official list & per-token rates 2026
Gemini is among the best value in frontier AI. Here are the current Gemini API prices per model, verified September 2026 — Google's official list alongside Kunavo's rates at about 70% less — with worked cost examples and the cheapest way to call Gemini in production.
Read guide - Pricing·9 min read
Anthropic Claude API pricing September 2026 (official rates)
Claude is the default for reasoning, coding and agents. Here are the current Claude API prices per model, verified September 2026 — Anthropic's official list alongside Kunavo's rates at 30–60% less — with worked examples and the levers that cut a Claude bill the most.
Read guide - API·7 min read
Claude API — the complete guide to calling Claude on Kunavo
Claude is Anthropic's frontier family for reasoning, coding and agents. This is the hub for calling it on Kunavo — keys, SDKs, models, pricing and caching, all linked.
Read guide - Video·7 min read
Sora API — text-to-video on an OpenAI-compatible workflow
Sora is OpenAI's text-to-video model. On Kunavo the live text-to-video model is Google Veo 3.1, on the same OpenAI-style video endpoint — here's the workflow, with Sora access on the roadmap.
Read guide - Cost·11 min read
AI cost optimization — the complete guide to cutting 70-90% off your LLM bill
If you're paying more than $1,000/month on LLM APIs, this guide will save you at least half. Every technique is published with measured savings and runnable code. Stack 3-5 of them and 70% reduction is realistic without quality tradeoff.
Read guide - Architecture·14 min read
RAG implementation guide — production retrieval-augmented generation with Claude and Gemini
Build a production RAG system that actually works — not just a 100-line demo. Covers chunking strategy, embedding model choice, retrieval ranking, prompt structure, hallucination control, and how to keep monthly costs in three figures even at 10,000 queries/day.
Read guide - Compliance·12 min read
AI compliance guide — GDPR, DSGVO, RGPD, LGPD, KVKK and Japan's tokutei-shoutorihiki
Shipping AI in production means meeting the data-protection regime of every country you sell to. This guide is the practical playbook — what each regulator actually checks, what Kunavo provides under our DPA, and what you still need to do on your side.
Read guide - Multimodal·10 min read
Multimodal AI guide — text, image, video and audio under one OpenAI-compatible API
Most aggregators only cover text. Kunavo gives you image (Nano Banana, GPT-Image), video (Veo 3.1), audio (Suno) on the same API. This guide is when to use which, and how to chain them for production-grade multimodal workflows.
Read guide