OpenRouter is the usual answer to “one API for every LLM”. It lists 500+ models from 80+ providers, routes each request to a provider by price, latency or uptime, and falls back automatically when one fails; it has since added image, video and speech endpoints. APIMODELS overlaps on the popular commercial LLMs (Claude, GPT, Gemini, DeepSeek, Grok, Qwen, GLM) and carries a larger media catalog behind the same key: 40 video models against the 29 in OpenRouter’s video API, plus 37 image and 21 audio models including Suno and ElevenLabs.
On LLMs the comparison is list price against discount. OpenRouter charges Anthropic’s $2 / $10 per million tokens for Claude Sonnet 5.5 and OpenAI’s $2 / $10 for GPT-6.1 Sol; APIMODELS charges $1.20 / $6 and $1 / $5. Gemini 3.1 Pro Preview is $2 / $12 there and $0.882 / $5.294 here. On video, Veo 3.1 (8 seconds, 1080p, with audio) is $1.00 here against $3.20, and Grok Imagine Video 1.5 at 720p is $0.0529 per second against $0.14. Prices were read from OpenRouter’s model pages and public API on 6 October 2026.
What you trade away is control. On OpenRouter you can pin providers, sort them by throughput or price, require zero data retention and bring your own provider keys; APIMODELS picks the upstream route for you. What you get instead is stability you can check plus a flat price: many models here have more than one upstream route with automatic failover, failed requests are never billed, every LLM response carries an x-apimodels-request-id header whose exact cost GET /v1/records/{id} returns, and media jobs retry their callback for up to 30 minutes.
The OpenRouter column comes from its pricing page, docs, model pages and public models API, checked on 2026-10-06; sources are at the end of this page.
| APIMODELS | OpenRouter | |
|---|---|---|
| What it is | Multi-model gateway with its own per-call and per-token prices | LLM router across providers at provider list price, plus image, video and speech APIs |
| Model coverage | 149 models: 49 LLM, 37 image, 40 video, 21 audio, 2 embedding (October 2026) | 500+ models from 80+ providers (site figures); 29 models in its video API |
| Pricing | Own prices, below list on most commercial LLMs; top-ups credit the full amount | Provider list price, "no markup on inference"; 5.5% fee on card credit purchases ($0.80 minimum), 5% on crypto |
| Failed requests | Never charged | Errored attempts not charged for model tokens; add-ons already run may be |
| LLM API | OpenAI-compatible /v1/chat/completions and /v1/responses, native Anthropic /v1/messages, native Gemini | OpenAI-compatible chat completions, Anthropic-compatible /api/v1/messages, stateless Responses API |
| Video API | Async: create, then poll or callback_url | Async: job with polling_url, or signed callback_url |
| Provider control | None exposed; automatic failover between our own routes | Per-request provider order, sort by price / throughput / latency, ZDR, BYOK |
| Payment | Card (Stripe), PayPal, Alipay, USDT; from $10 | Card, AliPay, USDC; invoicing and purchase orders on Enterprise |
| Free usage | $0.10 trial credit for consumer-email sign-ups | 25+ free models, 50 requests a day (1,000 after buying $10 of credit) |
USD, checked on 2026-10-06. OpenRouter prices are before its 5.5% card top-up fee. Nano Banana Pro is token-billed on OpenRouter at $120 per million image-output tokens; the per-image figure uses Google’s own conversion (1,120 tokens for a 1K or 2K image, 2,000 for 4K).
| Model and unit | APIMODELS | OpenRouter | Lower price |
|---|---|---|---|
| Claude Sonnet 5.5, per 1M input / output tokens | $1.20 / $6.00 | $2.00 / $10.00 | APIMODELS |
| Claude Opus 5.5, per 1M input / output tokens | $2.40 / $12.00 | $4.00 / $20.00 | APIMODELS |
| GPT-6.1 Sol, per 1M input / output tokens | $1.00 / $5.00 | $2.00 / $10.00 | APIMODELS |
| Gemini 3.1 Pro Preview, per 1M input / output tokens | $0.882 / $5.294 | $2.00 / $12.00 | APIMODELS |
| Nano Banana Pro, 1K–2K / 4K, per image | $0.08 / $0.13 | ≈$0.134 / ≈$0.24 | APIMODELS |
| Veo 3.1, 8 s, 1080p, with audio, per clip | $1.00 | $3.20 ($0.40/s) | APIMODELS |
| Grok Imagine Video 1.5, 720p / 1080p, per second | $0.0529 / $0.0882 | $0.14 / $0.25 | APIMODELS |
Choose OpenRouter when provider choice is the point. Claude Sonnet 5.5 is served there by five providers and DeepSeek V4.1 Flash by thirty; you can pin, order or exclude providers per request, sort them by price, throughput or latency, require zero data retention and bring your own provider keys. APIMODELS chooses the upstream for you and exposes none of those controls.
Choose OpenRouter for free models, open-weight breadth and enterprise controls. It lists 25+ free models and many open-weight models served by several providers each; its enterprise plans add SSO and SAML, SCIM, contractual SLAs and EU or US in-region routing.
fal.ai suits teams that also want to deploy their own model on serverless GPUs; its LLM endpoint is itself built on OpenRouter, so it is a media platform first.
Replicate suits teams that run open-source or custom models packaged with Cog; it lists Claude Sonnet 5 at Anthropic’s $2 / $10 but had no Claude 5.5 or GPT-6 listing on the check date.
Kie.ai suits teams that need its Suno V6 and Runway toolset and are fine with per-model-family endpoints. Each has its own comparison page, linked at the bottom.
For LLM code, switching from OpenRouter is a base URL and a model id. Point the OpenAI SDK at https://api.apimodels.app/v1 instead of https://openrouter.ai/api/v1, swap the key, and drop the vendor prefix: anthropic/claude-sonnet-5.5 becomes claude-sonnet-5-5, openai/gpt-6.1-sol becomes gpt-6.1-sol, google/gemini-3.1-pro-preview becomes gemini-3.1-pro-preview. GET https://api.apimodels.app/v1/models lists every id without a key.
Anthropic SDK users set base_url to https://api.apimodels.app (no /v1) and keep native tool use, thinking blocks and streaming events. OpenRouter-only request fields, such as provider preferences, the models fallback array and the :nitro or :floor suffixes, have no equivalent here, so remove them.
Video moves from OpenRouter’s POST /api/v1/videos to POST /v1/video/generations: send the model, prompt and settings, then poll GET /v1/video/generations?task_id=… or pass a callback_url. Results come back as resultUrls and stay downloadable for 30 days.
| Fact | Source |
|---|---|
| LLM prices | openrouter.ai/api/v1/models; model pages anthropic/claude-sonnet-5.5, anthropic/claude-opus-5.5, openai/gpt-6.1-sol, google/gemini-3.1-pro-preview |
| Nano Banana Pro token price | openrouter.ai/google/gemini-3-pro-image; per-image conversion from ai.google.dev/gemini-api/docs/pricing |
| Video prices and API | openrouter.ai/api/v1/videos/models; google/veo-3.1, x-ai/grok-imagine-video-1.5; docs video-generation guide |
| Fees, free models, payment, refunds | openrouter.ai/pricing; openrouter.ai/docs/faq; openrouter.ai/docs/api_reference/limits |
| Provider routing and provider counts | openrouter.ai/docs/guides/routing/provider-selection; each model page |
| Enterprise and in-region routing | openrouter.ai/pricing; openrouter.ai/docs/guides/features/in-region-routing |
| APIMODELS prices | apimodels.app/pricing and each model page |
cURL
# Before (OpenRouter): https://openrouter.ai/api/v1/chat/completions, model "anthropic/claude-sonnet-5.5"
# After (APIMODELS): same OpenAI request shape, new base URL, no vendor prefix
curl https://api.apimodels.app/v1/chat/completions \
-H "Authorization: Bearer $APIMODELS_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "claude-sonnet-5-5",
"messages": [{"role": "user", "content": "Summarise this changelog in three bullet points: ..."}]
}'
# the x-apimodels-request-id response header → GET /v1/records/{id} returns the exact chargePython
from openai import OpenAI
# Before: OpenAI(base_url="https://openrouter.ai/api/v1", api_key=OPENROUTER_KEY)
client = OpenAI(base_url="https://api.apimodels.app/v1", api_key="YOUR_APIMODELS_KEY")
for model in ["claude-sonnet-5-5", "gpt-6.1-sol", "gemini-3.1-pro-preview"]:
r = client.chat.completions.create(
model=model,
messages=[{"role": "user", "content": "One sentence: what is a vector database?"}],
)
print(model, "->", r.choices[0].message.content, r.usage)
# Images and video use the same key: POST /v1/images/generations, POST /v1/video/generationsFor the big commercial LLMs, yes, because OpenRouter charges provider list price and APIMODELS prices below it. Checked on 6 October 2026, per million input / output tokens: Claude Sonnet 5.5 is $2 / $10 on OpenRouter and $1.20 / $6 here; Claude Opus 5.5 is $4 / $20 against $2.40 / $12; GPT-6.1 Sol is $2 / $10 against $1 / $5; Gemini 3.1 Pro Preview is $2 / $12 against $0.882 / $5.294. OpenRouter also adds 5.5% when you buy credits by card. On video, Veo 3.1 (8 seconds, 1080p, with audio) is $1.00 here against $3.20, and Grok Imagine Video 1.5 at 720p is $0.0529 per second against $0.14; Nano Banana Pro is $0.08 per 1K–2K image against about $0.134.
OpenRouter’s reliability comes from breadth: it routes each model across several providers (five for Claude Sonnet 5.5) and falls back automatically when one errors. APIMODELS applies the same idea at a smaller scale: many models run on more than one upstream route with automatic failover, but you cannot see or choose the routes. What you can verify here: failed requests are never charged, every LLM response carries an x-apimodels-request-id header and GET /v1/records/{id} returns the exact amount billed, and image and video jobs retry their callback for up to 30 minutes. We do not publish an uptime percentage. If your application needs per-request provider control or zero-data-retention routing, that is an OpenRouter strength, not ours.
Change the base URL from https://openrouter.ai/api/v1 to https://api.apimodels.app/v1, swap the key, and drop the vendor prefix from model ids: anthropic/claude-sonnet-5.5 becomes claude-sonnet-5-5 and openai/gpt-6.1-sol becomes gpt-6.1-sol. The OpenAI SDK and most tools that accept a custom OpenAI base URL work unchanged; Anthropic SDK users point at https://api.apimodels.app instead and keep tool use, thinking blocks and streaming native. Remove OpenRouter-only fields such as provider preferences, the models fallback array and :nitro or :floor suffixes. Video jobs move from /api/v1/videos to /v1/video/generations, with a task_id to poll or a callback_url. The full model list is at GET /v1/models, no key required, so you can map ids before you switch.
Both take Alipay, which matters for teams in China. OpenRouter lists cards, AliPay and USDC, charges 5.5% on card credit purchases (minimum $0.80) and 5% on crypto, and offers invoicing and purchase orders on its Enterprise plan. APIMODELS takes cards through Stripe, PayPal, Alipay and USDT (TRC-20 or BEP-20); a top-up credits the full amount paid, from $10, and the packages from $50 up add a 2–5% bonus. Card payments are invoiced automatically, with your company name and tax number if you tick the business option at checkout. On refunds the policies differ: OpenRouter refunds unused credit within 24 hours of purchase, while purchased credit here is generally non-refundable, except for duplicate charges or billing errors.
No. OpenRouter lists 25+ free models, limited to 20 requests a minute and 50 a day, or 1,000 a day once you have bought at least $10 of credit, which is useful for prototypes and hobby projects. APIMODELS has no free models: new accounts on consumer email domains get $0.10 of trial credit, enough for four 1K GPT Image 2 images at medium quality or a few short Claude Sonnet 5.5 conversations, and everything after that is pay-as-you-go. If your workload runs entirely on free open-weight models, OpenRouter is the better fit. If it mixes paid LLMs with image or video generation, one balance here covers all of it.
Yes. apimodels.app and api.apimodels.app are reachable from mainland China, no organization verification is required, and one key covers Claude, GPT, Gemini, DeepSeek, Qwen and GLM on the OpenAI-compatible endpoint, plus the native Anthropic Messages endpoint that Claude Code uses. You can pay with Alipay at the order-time exchange rate or with USDT, and the site and docs are in Chinese. OpenRouter also accepts AliPay; whether its endpoints and its upstream providers are reachable from your network is something to test from your own servers before you commit.
