Releases·Tool Release
APIMart publishes AI API gateway pricing: Gemini 3 Pro $1.6 in/$9.6 out, GPT Image 2.5 $0.0085/image, Seedance 2.0 Mini $0.0106/sec
A set of multilingual GitHub repositories document APIMart, a third-party AI API gateway that fronts 307 models with a single OpenAI-compatible key. Published pricing (snapshot 2026-09-28) includes GPT Image 2.5 at $0.0085 per 1K image, Seedance 2.0 Mini at $0.01056 per 480P-second, and Qwen3.7 Flash input at $0.0228568/M tokens, with USD pay-as-you-go settlement and a $1 minimum top-up. The repos state the service has operated since last year with roughly 100,000 registered users, mostly enterprise accounts, and disclose no affiliation with any model vendor.
1714 engagement32 sourcesMon, Sep 28, 2026, 12:00 AM
Developments
Multilingual repos (Spanish, German, Indonesian, French) posted 2026-09-28 confirm 307 models live and describe reverse-engineered '-ext' route labeling.
Mon, Sep 28, 2026, 04:48 PM
AI Overview
- APIMart's GitHub docs list published gateway prices: GPT Image 2.5 (1K) $0.0085/image, Seedance 2.0 Mini (480P) $0.0106/second, Qwen3.7 Flash $0.0228568/M input tokens, with async submit-task_id-poll billing showing cost and credits_cost
- The gateway fronts 307 models (live /v1/models, 2026-09-28) through one OpenAI-compatible key and base URL, with $1 minimum top-up, USD pay-as-you-go settlement, and ~100,000 registered users
- APIMart documents itself as a third-party gateway unaffiliated with model vendors, noting console billing is authoritative
- APIMart lists Gemini 3 Pro at $1.6 input / $9.6 output, callable via curl with the gemini-3-pro-preview model id, task polling, and URL returned on completion
- APIMart publishes Portuguese-, Spanish-, French-, German-, Russian-, Japanese-, and Korean-language guides (llm-api-gateway-pt, cheapest-llm-api-es/-de, unofficial-api-gateway-ru, cheapest-llm-api-ja, nano-banana-reverse-api-ja, llm-api-gateway-free-tier-ko) alongside English pricing docs
- APIMart lists DeepSeek V4 Pro (deepseek-v4-pro) at $1.0286 input / $0.2057 cached input / $3.0857 output, callable via curl with task polling
- APIMart lists reverse-engineered routes alongside official ones, marked by an -ext model id suffix and covered by the same key
- APIMart publishes a Russian-language guide (unofficial-api-gateway-ru) showing image-generation curl calls and async task polling via task_id
Entities
Related
Products·launched
AI API gateway/relay: 300+ models behind one OpenAI-compatible key with published per-call pricing
An API gateway/relay (apimart.ai) that exposes 307 models — LLMs, image generation, video generation — through a single OpenAI-compatible base URL and key, with published unit prices and per-call cost shown in every response. It targets developers who want pay-as-you-go USD access to many models without VPNs, multiple accounts, or minimum commitments beyond a $1 top-up.
An API gateway/relay (apimart.ai) that exposes 307 models — LLMs, image generation, video generation — through a single OpenAI-compatible base URL and key, with published unit prices and per-call cost shown in every response. It targets developers who want pay-as-you-go USD access to many models without VPNs, multiple accounts, or minimum commitments beyond a $1 top-up.
050309 engagement5 sourcesgithubThu, Sep 24, 2026, 05:05 PM
Releases·Tool Release
APIMart relays GPT-Image-2.5 and 300+ models through one OpenAI-compatible endpoint from $1 minimum
Two GitHub posts document APIMart, a relay gateway exposing 300+ image, video, and language models behind a single OpenAI-compatible endpoint at api.apimart.ai/v1. Listed pricing includes image2.5 (GPT-Image-2.5) from $0.0085 per 1K image, Seedance 2.5 from $0.0961/sec, and cached LLM input from $0.40/M, with pay-as-you-go billing from a $1 minimum. The posts also ship a reproducible framework comparing four gateway archetypes — self-hosted open source, managed control plane, relayed per-unit route, and direct vendor API — scored on protocol compatibility, retry/idempotency semantics, and observability. For practitioners, the pitch is swapping models by changing the 'model' field without client changes.
Two GitHub posts document APIMart, a relay gateway exposing 300+ image, video, and language models behind a single OpenAI-compatible endpoint at api.apimart.ai/v1. Listed pricing includes image2.5 (GPT-Image-2.5) from $0.0085 per 1K image, Seedance 2.5 from $0.0961/sec, and cached LLM input from $0.40/M, with pay-as-you-go billing from a $1 minimum. The posts also ship a reproducible framework comparing four gateway archetypes — self-hosted open source, managed control plane, relayed per-unit route, and direct vendor API — scored on protocol compatibility, retry/idempotency semantics, and observability. For practitioners, the pitch is swapping models by changing the 'model' field without client changes.
98 engagement2 sourcesgithubWed, Sep 16, 2026, 07:13 AM
Social·Community
GLM-5.3 Flash pricing listed on APIMart: $0.06 input, $0.012 cached input, $0.20 output per unit via OpenAI-compatible gateway
A GitHub guide documents how to call GLM-5.3-Flash through the APIMart third-party gateway, listing observed per-unit pricing and code examples. It is vendor-independent documentation for developers accessing the model over an OpenAI-compatible API.
A GitHub guide documents how to call GLM-5.3-Flash through the APIMart third-party gateway, listing observed per-unit pricing and code examples. It is vendor-independent documentation for developers accessing the model over an OpenAI-compatible API.
56 engagement1 sourcegithubThu, Sep 24, 2026, 06:57 PM
Products·launched
OpenAI-compatible router pooling free LLM provider tiers behind one endpoint
A router that pools free tiers from multiple LLM providers (Groq, Cerebras, Gemini) behind a single OpenAI-compatible address. Scripts using /v1/chat/completions work unchanged; when one provider's rate limit is hit mid-run, requests fail over to the next provider. It solves rate-limit interruptions for people scripting against free LLM APIs.
A router that pools free tiers from multiple LLM providers (Groq, Cerebras, Gemini) behind a single OpenAI-compatible address. Scripts using /v1/chat/completions work unchanged; when one provider's rate limit is hit mid-run, requests fail over to the next provider. It solves rate-limit interruptions for people scripting against free LLM APIs.
1 engagement1 sourceredditMon, Sep 28, 2026, 07:31 PM
Social·Community
GPT Image 2.5 pricing is a tier problem, not a model problem, and 16:9 costs less than square
A Reddit user analyzed GPT Image 2.5 (Flare and Sunburst) pricing, showing both models share identical rates and that cost is driven by the quality step and output size, not model choice. Output is billed per token, producing a 36x cost spread within one canvas size, with the default 'high' step far above advertised low-step figures. The analysis also finds tall and 16:9 canvases cheaper than square, and behavioral differences between the two models when raising quality steps.
A Reddit user analyzed GPT Image 2.5 (Flare and Sunburst) pricing, showing both models share identical rates and that cost is driven by the quality step and output size, not model choice. Output is billed per token, producing a 36x cost spread within one canvas size, with the default 'high' step far above advertised low-step figures. The analysis also finds tall and 16:9 canvases cheaper than square, and behavioral differences between the two models when raising quality steps.
4 engagement1 sourceredditTue, Sep 22, 2026, 06:26 AM