# KovAPI - complete reference (llms-full.txt)

> KovAPI (formerly NezhaGate) is an OpenAI-compatible AI model gateway: one endpoint and one API key for GPT, Claude, Gemini, Grok, DeepSeek, Kimi, GLM and Qwen chat plus image and video generation (Seedance, Veo, Wan, MiniMax H3, Nano Banana, GPT Image). Pay as you go in USD; failed requests are never billed. Base URL: https://kovapi.com/v1. Last updated 2026-10-26.

Every price below is read live from the billing table at the moment this file is generated. Official list prices carry the date we checked them; a saving is shown only where we are below the official price. Measured figures come from real customer calls and appear only when the sample is large enough.

```
base_url: https://kovapi.com/v1
Authorization: Bearer <YOUR_API_KEY>
```

## Key facts
- API: OpenAI-compatible. POST https://kovapi.com/v1/chat/completions and /v1/responses for chat; /v1/images/generations and /v1/images/edits for images; /v1/videos/generations for video. Images and videos are async jobs: submit, then poll GET /v1/images/jobs/{id} or /v1/videos/jobs/{id}. Claude models are also served on the Anthropic Messages API at https://kovapi.com/anthropic/v1/messages.
- Billing: chat per 1M input / output tokens, cache reads at 10% of the input price; images per picture by resolution; video per second of output (Grok Imagine Video 1.5, Seedance 2.5 · Special, Seedance 2.0 · Clip, Veo 3.1 per clip). Failed requests cost nothing; image and video jobs hold credit at submit and refund it in full if they fail.
- Top-up: WeChat Pay from CNY 10, USDT and bank card from $5; $1 = 200 credits; credits never expire; unused top-ups are refundable within 7 days ($1.40 fee per refund for top-ups from 2026-10-29, $0.40 before). Larger packs add credit: $50: card/USDT/WeChat Pay +3%; $500: card/USDT +7%; $1,250: card +10%, USDT +12%.
- Availability: the public API answered 98.11% of external probes over the last 30 days (one probe every 2 minutes); live per-model status at https://kovapi.com/en/status.
- Automatic retries: a request that errors, times out or hits a capacity limit is retried automatically before any output is streamed, and is never billed twice.
- Rate limits: no default per-account limit; when a model is at capacity the API answers 429 with Retry-After. Per-key RPM, daily budget, total quota, model and IP allow-lists are self-service.
- Data: call logs stay 60 days in the console, their text content is deleted after 30 days, generated images and videos after 15 days, uploaded references after 7 days. Content is never used for training.
- Support: support@kovapi.com, Telegram https://t.me/+-erBoH9-AYY4MTM1 and WeChat; Beijing time 10:00-19:00, replies within 1 hour.

## Chat models (live prices, USD per 1M tokens)
- [GPT-5.6 Sol](https://kovapi.com/en/model/gpt-5.6-sol): OpenAI. input $1.20 / output $6.00, cache read $0.12. official $4.00 / $20.00 (vendor list price, checked 2026-10-02), save 70%. Frontier reasoning, agentic coding, long-horizon tasks, structured output. ~272K tokens (GPT-5 family context window; system-prompt overhead counts toward it). measured on real customer calls over the last 7 days: 98.3% success. first token p50 6.2 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gpt-5.6-sol.
- [GPT-5.6 Luna](https://kovapi.com/en/model/gpt-5.6-luna): OpenAI. input $0.07 / output $0.40, cache read $0.007. official $0.20 / $1.20 (vendor list price, checked 2026-10-02), save 65% / 66%. High-speed chat, agentic coding, high-volume low-latency, structured output. ~272K tokens (GPT-5 family context window; system-prompt overhead counts toward it). measured on real customer calls over the last 30 days: 92.7% success. docs: https://docs.kovapi.com/en/gpt-5.6-luna.
- [GPT-6 Astra](https://kovapi.com/en/model/gpt-6-astra): OpenAI. input $3.00 / output $15.00, cache read $0.30. official $10.00 / $50.00 (vendor list price, checked 2026-10-02), save 70%. Deep reasoning, agentic coding, very long context, image input, structured output. 1,050,000 tokens (OpenAI published figure); up to 128,000 output tokens per call. System-prompt overhead counts toward it. measured on real customer calls over the last 7 days: 99.2% success. first token p50 7.4 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gpt-6-astra.
- [GPT-6.1 Sol](https://kovapi.com/en/model/gpt-6.1-sol): OpenAI. input $0.60 / output $3.00, cache read $0.06. official $2.00 / $10.00 (vendor list price, checked 2026-10-02), save 70%. Agentic coding, complex coding, professional document and analysis work, reasoning, very long context, image input, structured output. 1,050,000 tokens (OpenAI published figure); up to 128,000 output tokens per call. System-prompt overhead counts toward it. measured on real customer calls over the last 7 days: 91.2% success. first token p50 8.3 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gpt-6.1-sol.
- [GPT-6 Sol](https://kovapi.com/en/model/gpt-6-sol): OpenAI. input $0.60 / output $3.00, cache read $0.06. official $2.00 / $10.00 (vendor list price, checked 2026-10-02), save 70%. Complex coding, agentic workflows, reasoning, very long context, image input, structured output. 1,050,000 tokens (OpenAI published figure); up to 128,000 output tokens per call. System-prompt overhead counts toward it. measured on real customer calls over the last 7 days: 94.2% success. first token p50 4.8 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gpt-6-sol.
- [GPT-6 Luna](https://kovapi.com/en/model/gpt-6-luna): OpenAI. input $0.07 / output $0.40, cache read $0.007. official $0.10 / $0.50 (vendor list price, checked 2026-10-02), save 30% / 20%. High-volume chat, low-latency agent steps, classification and extraction, very long context, structured output. 1,050,000 tokens (OpenAI published figure); up to 128,000 output tokens per call. System-prompt overhead counts toward it. measured on real customer calls over the last 7 days: 93.7% success. first token p50 3.2 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gpt-6-luna.
- [Claude Sonnet 4.6](https://kovapi.com/en/model/claude-sonnet-4-6): Anthropic. input $1.50 / output $7.50, cache read $0.15. official $3.00 / $15.00 (vendor list price, checked 2026-10-02), save 50%. Chat, code, reasoning, long context. ~200K tokens. measured on real customer calls over the last 7 days: 100% success. first token p50 3.1 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/claude-sonnet-4-6.
- [Claude Opus 5](https://kovapi.com/en/model/claude-opus-5): Anthropic. input $4.00 / output $20.00, cache read $0.40. official $5.00 / $25.00 (vendor list price, checked 2026-10-02), save 20%. Deep reasoning, code, agents, long context, image understanding. ~200K tokens. first token p50 6.2 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/claude-opus-5.
- [Claude Fable 5](https://kovapi.com/en/model/claude-fable-5): Anthropic. input $8.00 / output $40.00, cache read $0.80. official $10.00 / $50.00 (vendor list price, checked 2026-10-02), save 20%. Chinese writing, narrative creation, long-form generation, chat, code, image understanding. ~200K tokens. first token p50 6.2 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/claude-fable-5.
- [Claude Opus 5.5](https://kovapi.com/en/model/claude-opus-5-5): Anthropic. input $2.50 / output $12.50, cache read $0.125. official $4.00 / $20.00 (vendor list price, checked 2026-10-02), save 37%. Complex reasoning, long-horizon agentic coding, code engineering, tool calling, image and PDF understanding. 1,000,000 tokens (Anthropic published figure); up to 128,000 output tokens per request. measured on real customer calls over the last 30 days: 89.1% success. docs: https://docs.kovapi.com/en/claude-opus-5-5.
- [Claude Sonnet 5.5](https://kovapi.com/en/model/claude-sonnet-5-5): Anthropic. input $1.50 / output $7.50, cache read $0.15. official $2.00 / $10.00 (vendor list price, checked 2026-10-02), save 25%. Everyday coding and agents, long context, tool calling, image and PDF understanding, production traffic at scale. 1,000,000 tokens (Anthropic published figure); up to 128,000 output tokens per request. measured on real customer calls over the last 7 days: 98.5% success. docs: https://docs.kovapi.com/en/claude-sonnet-5-5.
- [Gemini 4 Argon](https://kovapi.com/en/model/gemini-4-argon): Google. input $0.50 / output $3.00, cache read $0.05. official $2.00 / $10.00 (vendor list price, checked 2026-10-02), save 75% / 70%. Beta · reasoning · long context. ~1,000,000 tokens (million-token long context). measured on real customer calls over the last 30 days: 52.1% success. docs: https://docs.kovapi.com/en/gemini-4-argon.
- [Gemini 3.1 Pro](https://kovapi.com/en/model/gemini-3.1-pro): Google. input $0.50 / output $3.00, cache read $0.05. official $2.00 / $12.00 (vendor list price, checked 2026-10-02), save 75%. Chat, reasoning, very long context, multimodal. ~1,000,000 tokens (million-token long context). measured on real customer calls over the last 30 days: 99.7% success. docs: https://docs.kovapi.com/en/gemini-3.1-pro.
- [Gemini 3.8 Flash](https://kovapi.com/en/model/gemini-3.8-flash): Google. input $0.60 / output $3.60, cache read $0.06. official $0.75 / $3.75 (vendor list price, checked 2026-10-02), save 20% / 4%. Chat, reasoning, adaptive thinking, image input, very long context. ~1,000,000 tokens (million-token long context). measured on real customer calls over the last 7 days: 100% success. first token p50 10.1 s (weekly test, 2026-10-27). docs: https://docs.kovapi.com/en/gemini-3.8-flash.
- [DeepSeek V4.1 Flash](https://kovapi.com/en/model/deepseek-v4.1-flash): DeepSeek. input $0.075 / output $0.30, cache read $0.0015. official $0.15 / $0.60 (vendor list price, checked 2026-10-02), save 50%. Chat, reasoning, coding, tool calling, switchable thinking mode. measured on real customer calls over the last 7 days: 99.6% success. docs: https://docs.kovapi.com/en/deepseek-v4.1-flash.
- [DeepSeek V4 Flash 0731](https://kovapi.com/en/model/deepseek-v4-flash-0731): DeepSeek. input $0.13 / output $0.51, cache read $0.004. official $0.15 / $0.60 (vendor list price, checked 2026-10-02), save 13% / 15%. Chat, reasoning, coding, tool calling, switchable thinking mode, pinned version. measured on real customer calls over the last 30 days: 99.5% success. docs: https://docs.kovapi.com/en/deepseek-v4-flash-0731.
- [GLM-5.3](https://kovapi.com/en/model/glm-5.3): Z.ai. input $1.12 / output $3.52, cache read $0.21. official $1.40 / $4.40 (vendor list price, checked 2026-10-02), save 20%. Chat, coding, agentic workflows, reasoning, Chinese and English writing, tool calling. measured on real customer calls over the last 7 days: 100% success. docs: https://docs.kovapi.com/en/glm-5.3.
- [Qwen3.8 Max](https://kovapi.com/en/model/qwen3.8-max): Alibaba. input $1.40 / output $4.20, cache read $0.175. official $2.00 / $6.00 (vendor list price, checked 2026-10-02), save 30%. Reasoning, coding, agents and tool calling, Chinese and English writing, image understanding. docs: https://docs.kovapi.com/en/qwen3.8-max.
- [Qwen3.8 Flash](https://kovapi.com/en/model/qwen3.8-flash): Alibaba. input $0.10 / output $0.33, cache read $0.012. official $0.15 / $0.47 (vendor list price, checked 2026-10-02), save 33% / 29%. High-volume chat, classification and extraction, summarising and rewriting, image understanding and OCR, tool calling. measured on real customer calls over the last 30 days: 99.1% success. docs: https://docs.kovapi.com/en/qwen3.8-flash.
- [Grok 4.7](https://kovapi.com/en/model/grok-4.7): xAI. input $0.30 / output $0.90, cache read $0.075. official $2.00 / $6.00 (vendor list price, checked 2026-10-02), save 85%. Complex reasoning, coding, long-document analysis, tool calling, structured output, image understanding. measured on real customer calls over the last 7 days: 87.9% success. docs: https://docs.kovapi.com/en/grok-4.7.

## Image models (live prices, USD per image)
- [GPT Image 2](https://kovapi.com/en/model/gpt-image-2): OpenAI. $0.005 (1K) / $0.025 (2K) / $0.04 (4K) per image. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 99.8% success, median 23 s per image (p90 40 s). providers compared: https://kovapi.com/en/compare/gpt-image. docs: https://docs.kovapi.com/en/gpt-image-2.
- [GPT Image 2.5 Flare](https://kovapi.com/en/model/gpt-image-2.5-flare): OpenAI. $0.005 (1K) / $0.03 (2K) / $0.05 (4K) per image. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 99.8% success, median 24 s per image (p90 39 s). providers compared: https://kovapi.com/en/compare/gpt-image. docs: https://docs.kovapi.com/en/gpt-image-2.5-flare.
- [GPT Image 2.5 Sunburst](https://kovapi.com/en/model/gpt-image-2.5-sunburst): OpenAI. $0.005 (1K) / $0.03 (2K) / $0.05 (4K) per image. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 99.2% success, median 26 s per image (p90 43 s). docs: https://docs.kovapi.com/en/gpt-image-2.5-sunburst.
- [Nano Banana 2](https://kovapi.com/en/model/nano-banana-2): Google. $0.015 (1K) / $0.025 (2K) / $0.04 (4K) per image. official $0.067 / $0.101 / $0.151 (vendor list price, checked 2026-10-02), save 77% / 75% / 73%. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 100% success, median 13 s per image (p90 30 s). providers compared: https://kovapi.com/en/compare/nano-banana. docs: https://docs.kovapi.com/en/nano-banana-2.
- [Nano Banana Pro](https://kovapi.com/en/model/nano-banana-pro): Google. $0.02 (1K) / $0.03 (2K) / $0.05 (4K) per image. official $0.134 / $0.134 / $0.24 (vendor list price, checked 2026-10-02), save 85% / 77% / 79%. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 100% success, median 34 s per image (p90 62 s). providers compared: https://kovapi.com/en/compare/nano-banana. docs: https://docs.kovapi.com/en/nano-banana-pro.
- [Nano Banana 2.1](https://kovapi.com/en/model/nano-banana-2.1): Google. $0.015 (1K) / $0.025 (2K) / $0.04 (4K) per image. official $0.0336 / $0.0504 / $0.0756 (vendor list price, checked 2026-10-02), save 55% / 50% / 47%. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 100% success, median 30 s per image (p90 58 s). docs: https://docs.kovapi.com/en/nano-banana-2.1.
- [Grok Imagine Image](https://kovapi.com/en/model/grok-imagine-image): xAI. $0.005 (1K) / $0.005 (2K) per image. official $0.02 / $0.02 (vendor list price, checked 2026-10-02), save 75%. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 98% success, median 15 s per image (p90 20 s). docs: https://docs.kovapi.com/en/grok-imagine-image.
- [Grok Imagine Image Quality](https://kovapi.com/en/model/grok-imagine-image-quality): xAI. $0.01 (1K) / $0.01 (2K) per image. official $0.05 / $0.07 (vendor list price, checked 2026-10-02), save 80% / 85%. text-to-image and image editing (/v1/images/edits). measured on real customer calls over the last 7 days: 91.2% success, median 14 s per image (p90 18 s). docs: https://docs.kovapi.com/en/grok-imagine-image-quality.

## Video models (live prices, USD)
- [Veo 3.1](https://kovapi.com/en/model/veo-3.1): Google. 720p $0.20 per clip, 1080p $0.22 per clip. fixed 8 s per clip. up to 3 reference images. first frame, first + last frame, or subject references. native audio always on. providers compared: https://kovapi.com/en/compare/veo. docs: https://docs.kovapi.com/en/veo-3.1.
- [Gemini Omni Flash](https://kovapi.com/en/model/gemini-omni-flash): Google. 720P $0.05/s, 1080P $0.05/s. 6 / 8 / 10 s per clip. up to 3 reference images. first frame, first + last frame, or subject references. prompt up to 20,000 characters. native audio always on. providers compared: https://kovapi.com/en/compare/veo. docs: https://docs.kovapi.com/en/gemini-omni-flash.
- [Seedance 2.5 · Token](https://kovapi.com/en/model/seedance-2.5-token): ByteDance. images are references (@Image1, @Image2 ...), no first-frame control. docs: https://docs.kovapi.com/en/seedance-2.5-token.
- [Seedance 2.5 · Token Draft](https://kovapi.com/en/model/seedance-2.5-token-draft): ByteDance. images are references (@Image1, @Image2 ...), no first-frame control. docs: https://docs.kovapi.com/en/seedance-2.5-token-draft.
- [Seedance 2.5](https://kovapi.com/en/model/seedance-2.5): ByteDance. 480P $0.10/s, 720P $0.158/s, 1080P $0.33/s. official 480P $0.103/s, 720P $0.231/s (vendor list price, checked 2026-10-02), save 480P 2%, 720P 31%. 4-30 s per clip. up to 30 reference images. 10 reference videos (30 s in total). 10 reference audio clips. images are references (@Image1, @Image2 ...), no first-frame control. prompt up to 11,000 characters. native audio always on. reference videos are not billed. measured on real customer calls over the last 7 days: 84.5% success, median 8 min per clip (p90 20.1 min). providers compared: https://kovapi.com/en/compare/seedance. docs: https://docs.kovapi.com/en/seedance-2.5.
- [Seedance 2.0](https://kovapi.com/en/model/seedance-2.0): ByteDance. 720P $0.15/s, 1080P $0.30/s (480P coming soon). official 720P $0.151/s, 1080P $0.374/s (vendor list price, checked 2026-10-02), save 1080P 19%. 5-15 s per clip. up to 9 reference images. 3 reference videos (15 s in total). 3 reference audio clips. images are references (@Image1, @Image2 ...), no first-frame control. native audio always on. reference videos are not billed. measured on real customer calls over the last 30 days: 64.4% success, median 4.8 min per clip (p90 7.2 min). providers compared: https://kovapi.com/en/compare/seedance. docs: https://docs.kovapi.com/en/seedance-2.0.
- [Seedance 2.0 Fast](https://kovapi.com/en/model/seedance-2.0-fast): ByteDance. 480P $0.046/s, 720P $0.075/s. official 480P $0.054/s, 720P $0.121/s (vendor list price, checked 2026-10-02), save 480P 14%, 720P 38%. 5-15 s per clip. up to 9 reference images. 3 reference videos (15 s in total). 3 reference audio clips. images are references (@Image1, @Image2 ...), no first-frame control. native audio always on. reference videos are not billed. providers compared: https://kovapi.com/en/compare/seedance. docs: https://docs.kovapi.com/en/seedance-2.0-fast.
- [Seedance 2.0 Mini](https://kovapi.com/en/model/seedance-2.0-mini): ByteDance. 480P $0.022/s, 720P $0.046/s. 5-15 s per clip. up to 9 reference images. 3 reference videos (15 s in total). 3 reference audio clips. images are references (@Image1, @Image2 ...), no first-frame control. native audio always on. reference videos are not billed. docs: https://docs.kovapi.com/en/seedance-2.0-mini.
- [Seedance 2.0 · Clip](https://kovapi.com/en/model/seedance-2.0-clip): ByteDance. 720P $1.00 per clip. 5 / 10 / 15 s per clip. up to 9 reference images. images are references (@Image1, @Image2 ...), no first-frame control. native audio always on. docs: https://docs.kovapi.com/en/seedance-2.0-clip.
- [Seedance 2.5 · Special](https://kovapi.com/en/model/seedance-2.5-30s): ByteDance. 720P $2.00 per clip. official $6.93 per clip (vendor list price, checked 2026-10-02), save 71%. 4-30 s per clip. up to 30 reference images. 10 reference audio clips. images are references (@Image1, @Image2 ...), no first-frame control. native audio always on. measured on real customer calls over the last 7 days: 65% success, median 12.8 min per clip (p90 63.5 min). providers compared: https://kovapi.com/en/compare/seedance. docs: https://docs.kovapi.com/en/seedance-2.5-30s.
- [Wan 3.0](https://kovapi.com/en/model/wan3.0-video): Alibaba. 480P $0.04/s, 720P $0.065/s, 1080P $0.13/s. official 480P $0.05/s, 720P $0.10/s, 1080P $0.20/s (vendor list price, checked 2026-10-02), save 480P 20%, 720P 35%, 1080P 35%. 4-30 s per clip. up to 10 reference images. 5 reference videos (15 s in total). first frame, first + last frame, or subject references. no audio switch. reference-video seconds are billed. measured on real customer calls over the last 30 days: 60.9% success, median 3.5 min per clip (p90 7.2 min). providers compared: https://kovapi.com/en/compare/wan. docs: https://docs.kovapi.com/en/wan3.0-video.
- [Wan 3.0 Prime](https://kovapi.com/en/model/wan3.0-video-prime): Alibaba.  (480P coming soon, 720P coming soon, 1080P coming soon). 4-30 s per clip. up to 10 reference images. 5 reference videos (15 s in total). first frame, first + last frame, or subject references. no audio switch. reference-video seconds are billed. providers compared: https://kovapi.com/en/compare/wan. docs: https://docs.kovapi.com/en/wan3.0-video-prime.
- [MiniMax H3](https://kovapi.com/en/model/minimax-h3): MiniMax. 1080P $0.036/s. official 768P $0.08/s, 2K $0.13/s (vendor list price, checked 2026-10-02; no 1080P tier), save 55% against 768P. 5-15 s per clip. up to 5 reference images. 3 reference audio clips. first-frame mode by default; image_role=reference for multi-image reference. measured on real customer calls over the last 30 days: 92% success, median 4.6 min per clip (p90 13.6 min). providers compared: https://kovapi.com/en/compare/minimax-h3. docs: https://docs.kovapi.com/en/minimax-h3.
- [Grok Imagine Video 1.5](https://kovapi.com/en/model/grok-imagine-video-1.5): xAI. 480P $0.30 per clip, 720P $0.40 per clip. 4-15 s per clip. up to 7 reference images. native audio always on. measured on real customer calls over the last 7 days: 63.2% success, median 81 s per clip (p90 101 s). docs: https://docs.kovapi.com/en/grok-imagine-video-1.5.

## Coming soon (not callable yet)
- [Claude Haiku 5.5](https://kovapi.com/en/model/claude-haiku-5-5): coming soon; Claude Haiku 5.5 API: Anthropic's fastest, cheapest and most capable small model yet, with a 1M-token context, the first Haiku with effort control, and list prices up to 90% below Haiku 4.5
- [Claude Fable 5.1](https://kovapi.com/en/model/claude-fable-5-1): coming soon; Anthropic's most capable widely released model: the successor to Fable 5, with a 1M-token context

## Price comparisons
- [Seedance API Pricing Compared: 2.5, 2.0, Fast](https://kovapi.com/en/compare/seedance): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)
- [Nano Banana 2 & Pro API Pricing Compared](https://kovapi.com/en/compare/nano-banana): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)
- [GPT Image 2 & 2.5 API Pricing Compared](https://kovapi.com/en/compare/gpt-image): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)
- [MiniMax H3 (Hailuo 3) API Pricing Compared](https://kovapi.com/en/compare/minimax-h3): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)
- [Veo 3.1 & Gemini Omni Flash API Pricing Compared](https://kovapi.com/en/compare/veo): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)
- [Wan 3.0 & Wan 3.0 Prime API Pricing Compared](https://kovapi.com/en/compare/wan): official list price, OpenRouter, fal, Kie, Replicate, APIMart and KovAPI side by side (checked 2026-10-03)

## Guides
- [Seedance 2.5 Special Tier: A Practical Guide](https://kovapi.com/en/learn/seedance-special-tier-guide): Use the Seedance 2.5 special tier for 30-second 720P videos: real samples, current pricing, prompt advice, local demand windows and the 120-minute refund policy.
- [Seedance 2.5 API: Price, Parameters and Code](https://kovapi.com/en/learn/seedance-2-5-api): Call the Seedance 2.5 video API: $0.158 per second at 720P, 4 to 30 seconds, up to 30 reference images, 10 videos and 10 audio clips, with copy-paste curl, Python and Node.js code.
- [Seedance 30-Second Video API: Price, Samples, Code](https://kovapi.com/en/learn/seedance-30-second-video): Seedance 2.5 renders 30 seconds in one shot. Two ways on KovAPI: seedance-2.5-30s at $2.00 per 720P clip, or seedance-2.5 per second. Samples with full prompts, prompt structure and code.

## Docs and machine-readable files
- [Docs home](https://docs.kovapi.com/en/): Docs home: API reference, user guide, integrations
- [API reference](https://docs.kovapi.com/en/api/md): Every endpoint, auth, async jobs, the error format and rate limits (Markdown)
- [API docs and integration guide](https://docs.kovapi.com/en/guide.md): The full user guide (Markdown)
- [Error codes](https://docs.kovapi.com/en/guide/errors/md): One error format for everything, what each code means, and what to do when you see it.
- [Async jobs](https://docs.kovapi.com/en/guide/async-jobs/md): How to submit image and video jobs, poll or get them pushed, download the result, read every status and field, and how refunds work.
- [Billing and refunds](https://docs.kovapi.com/en/guide/billing/md): How chat, images and video are billed, how holds settle, when nothing is charged, and how to check spend, top up and get refunds.
- [Rate limits and retries](https://docs.kovapi.com/en/guide/rate-limits/md): Account and per-key limits, what you get when a model is at capacity, and how to retry and set timeouts.
- [Authentication and API keys](https://docs.kovapi.com/en/guide/api-keys/md): Create, use, limit, rotate and protect your API keys.
- [Integrations](https://docs.kovapi.com/en/integrations): Setup for Claude Code, Cline, Cherry Studio and other tools
- [Docs MCP server](https://kovapi.com/mcp): A read-only MCP endpoint (Streamable HTTP, no key): search_docs, read_doc, list_models
- [Pricing JSON: machine-readable sell prices (New API compatible)](https://kovapi.com/api/pricing)
- [Agent skill for Claude Code / Codex: npx skills add https://kovapi.com](https://docs.kovapi.com/en/integrations/agent-skills)
- [Client setup guides (SillyTavern, Cherry Studio, Cline)](https://docs.kovapi.com/en/guides)

## Price table

| Model ID | Name | Type | KovAPI price | Official (in / out) | Vendor |
| --- | --- | --- | --- | --- | --- |
| gpt-5.6-sol | GPT-5.6 Sol | chat / reasoning | $1.20 / $6.00 | $4.00 / $20.00 | OpenAI |
| gpt-5.6-luna | GPT-5.6 Luna | chat / reasoning | $0.07 / $0.40 | $0.20 / $1.20 | OpenAI |
| gpt-6-astra | GPT-6 Astra | chat / reasoning | $3.00 / $15.00 | $10.00 / $50.00 | OpenAI |
| gpt-6.1-sol | GPT-6.1 Sol | chat / reasoning | $0.60 / $3.00 | $2.00 / $10.00 | OpenAI |
| gpt-6-sol | GPT-6 Sol | chat / reasoning | $0.60 / $3.00 | $2.00 / $10.00 | OpenAI |
| gpt-6-luna | GPT-6 Luna | chat / reasoning | $0.07 / $0.40 | $0.10 / $0.50 | OpenAI |
| gpt-image-2 | GPT Image 2 | image generation | $0.005 / $0.025 / $0.04 per image | - | OpenAI |
| gpt-image-2.5-flare | GPT Image 2.5 Flare | image generation | $0.005 / $0.03 / $0.05 per image | - | OpenAI |
| gpt-image-2.5-sunburst | GPT Image 2.5 Sunburst | image generation | $0.005 / $0.03 / $0.05 per image | - | OpenAI |
| nano-banana-2 | Nano Banana 2 | image generation | $0.015 / $0.025 / $0.04 per image | - | Google |
| nano-banana-pro | Nano Banana Pro | image generation | $0.02 / $0.03 / $0.05 per image | - | Google |
| nano-banana-2.1 | Nano Banana 2.1 | image generation | $0.015 / $0.025 / $0.04 per image | - | Google |
| grok-imagine-image | Grok Imagine Image | image generation | $0.005 / $0.005 per image | - | xAI |
| grok-imagine-image-quality | Grok Imagine Image Quality | image generation | $0.01 / $0.01 per image | - | xAI |
| claude-sonnet-4-6 | Claude Sonnet 4.6 | chat / reasoning | $1.50 / $7.50 | $3.00 / $15.00 | Anthropic |
| claude-opus-5 | Claude Opus 5 | chat / reasoning | $4.00 / $20.00 | $5.00 / $25.00 | Anthropic |
| claude-fable-5 | Claude Fable 5 | chat / reasoning | $8.00 / $40.00 | $10.00 / $50.00 | Anthropic |
| claude-opus-5-5 | Claude Opus 5.5 | chat / reasoning | $2.50 / $12.50 | $4.00 / $20.00 | Anthropic |
| claude-sonnet-5-5 | Claude Sonnet 5.5 | chat / reasoning | $1.50 / $7.50 | $2.00 / $10.00 | Anthropic |
| gemini-4-argon | Gemini 4 Argon | chat / reasoning | $0.50 / $3.00 | $2.00 / $10.00 | Google |
| gemini-3.1-pro | Gemini 3.1 Pro | chat / reasoning | $0.50 / $3.00 | $2.00 / $12.00 | Google |
| gemini-3.8-flash | Gemini 3.8 Flash | chat / reasoning | $0.60 / $3.60 | $0.75 / $3.75 | Google |
| deepseek-v4.1-flash | DeepSeek V4.1 Flash | chat / reasoning | $0.075 / $0.30 | $0.15 / $0.60 | DeepSeek |
| deepseek-v4-flash-0731 | DeepSeek V4 Flash 0731 | chat / reasoning | $0.13 / $0.51 | $0.15 / $0.60 | DeepSeek |
| glm-5.3 | GLM-5.3 | chat / reasoning | $1.12 / $3.52 | $1.40 / $4.40 | Z.ai |
| qwen3.8-max | Qwen3.8 Max | chat / reasoning | $1.40 / $4.20 | $2.00 / $6.00 | Alibaba |
| qwen3.8-flash | Qwen3.8 Flash | chat / reasoning | $0.10 / $0.33 | $0.15 / $0.47 | Alibaba |
| grok-4.7 | Grok 4.7 | chat / reasoning | $0.30 / $0.90 | $2.00 / $6.00 | xAI |
| veo-3.1 | Veo 3.1 | video generation | 720p $0.20 per clip, 1080p $0.22 per clip | - | Google |
| gemini-omni-flash | Gemini Omni Flash | video generation | 720P $0.05/s, 1080P $0.05/s | - | Google |
| seedance-2.5-token | Seedance 2.5 · Token | video generation | - | - | ByteDance |
| seedance-2.5-token-draft | Seedance 2.5 · Token Draft | video generation | - | - | ByteDance |
| seedance-2.5 | Seedance 2.5 | video generation | 480P $0.10/s, 720P $0.158/s, 1080P $0.33/s | - | ByteDance |
| seedance-2.0 | Seedance 2.0 | video generation | 720P $0.15/s, 1080P $0.30/s | - | ByteDance |
| seedance-2.0-fast | Seedance 2.0 Fast | video generation | 480P $0.046/s, 720P $0.075/s | - | ByteDance |
| seedance-2.0-mini | Seedance 2.0 Mini | video generation | 480P $0.022/s, 720P $0.046/s | - | ByteDance |
| seedance-2.0-clip | Seedance 2.0 · Clip | video generation | 720P $1.00 per clip | - | ByteDance |
| seedance-2.5-30s | Seedance 2.5 · Special | video generation | 720P $2.00 per clip | - | ByteDance |
| wan3.0-video | Wan 3.0 | video generation | 480P $0.04/s, 720P $0.065/s, 1080P $0.13/s | - | Alibaba |
| wan3.0-video-prime | Wan 3.0 Prime | video generation | - | - | Alibaba |
| minimax-h3 | MiniMax H3 | video generation | 1080P $0.036/s | - | MiniMax |
| grok-imagine-video-1.5 | Grok Imagine Video 1.5 | video generation | 480P $0.30 per clip, 720P $0.40 per clip | - | xAI |

## Code samples

### Chat (OpenAI SDK)

```python
from openai import OpenAI

client = OpenAI(
    base_url="https://kovapi.com/v1",
    api_key="YOUR_KOVAPI_API_KEY",
)

stream = client.chat.completions.create(
    model="gpt-6-astra",
    messages=[{"role": "user", "content": "Explain rate limiting in two sentences."}],
    stream=True,
)
for chunk in stream:
    if chunk.choices and chunk.choices[0].delta.content:
        print(chunk.choices[0].delta.content, end="", flush=True)
```

### curl

```bash
curl https://kovapi.com/v1/chat/completions \
  -H "Authorization: Bearer $KOVAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-6-astra",
    "messages": [{"role": "user", "content": "Explain rate limiting in two sentences."}],
    "stream": true
  }'
```

### Image (async job)

```bash
# 1) submit: returns HTTP 202 and a job id
curl https://kovapi.com/v1/images/generations \
  -H "Authorization: Bearer $KOVAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "gpt-image-2", "prompt": "A red fox in fresh snow, soft morning light", "size": "16:9", "resolution": "1K"}'

# 2) poll every 3 s until "status" is "succeeded", then read data[0].url
curl https://kovapi.com/v1/images/jobs/JOB_ID \
  -H "Authorization: Bearer $KOVAPI_API_KEY"
```

### Video (async job)

```python
import time
import requests

BASE = "https://kovapi.com/v1"
HEADERS = {"Authorization": "Bearer YOUR_KOVAPI_API_KEY"}

job = requests.post(f"{BASE}/videos/generations", headers=HEADERS, json={
    "model": "seedance-2.5",
    "prompt": "A paper boat drifting down a rainy street at night, neon reflections",
    "seconds": 5,
    "resolution": "720P",
    "size": "16:9",
}).json()

while True:
    r = requests.get(f"{BASE}/videos/jobs/{job['id']}", headers=HEADERS).json()
    if r["status"] == "succeeded":
        print(r["data"][0]["url"])
        break
    if r["status"] == "failed":
        raise RuntimeError(r["error"]["message"])
    time.sleep(15)
```

## Seedance content requirements

Every Seedance job goes through content review: the prompt, the reference files and the finished clip, picture and sound alike. A job that fails review ends with `content_policy` and is not charged.

### When review happens

The prompt and references are reviewed at submit; the finished clip is reviewed again, picture and sound, once it is generated. A rejection can therefore arrive seconds after submitting or only when rendering ends, which can take over half an hour. A job that ends with `status` `failed` and `error.code` `content_policy` failed review; with a webhook you receive a `video.failed` event.

### Not charged

A rejected job costs nothing: the amount held at submit is refunded in full when the job ends, and the call log shows it as not billed.

### Sound

The soundtrack (dialogue, narration, singing and effects) is generated by the model from your prompt and reviewed with the picture: spoken lines, narration and lyrics in the prompt are voiced or sung. For a clip without voices, leave lines and lyrics out of the prompt. Every Seedance clip comes with a soundtrack; it cannot be turned off. Uploaded reference audio is reviewed too, and copyrighted music used as reference audio may be rejected. When the sound fails review, `error.param` is `audio` and the message says whether it was the generated sound or your reference audio.

### Commonly rejected content

- Sexual content, nudity or sexual suggestion, including AI-generated or composited images
- Gore, cruelty, self-harm, and violent injury such as gunshot wounds and bleeding
- Firearms, explosives and drugs
- Any sexualised content involving minors; minors shown with alcohol, tobacco, drugs or in dangerous situations are rejected too
- Politicians, celebrities, athletes, business figures and other public figures, in looks or voice, including look-alikes
- Copyrighted characters and IP, such as Disney or Marvel characters
- Hateful or discriminatory content, and fake news

### Long storyboard scripts

One shot that breaks the rules fails the whole clip, and long storyboard scripts are where it happens most: a gunfight or a close-up of bleeding in an action scene, or minors in a drinking or smoking scene, sink the whole clip. Read the script through before submitting and remove or rewrite such shots.

### Real people

To use a photo of a real person as a reference you need the right to use their likeness.

### After a rejection

Resubmitting the same input is rejected again. Find the part that is likely at fault, reword the prompt or replace the reference, then submit again.

Error code: `content_policy` · `status: failed`

`upstream_line_unavailable`: If the generation service is overloaded or available capacity is exhausted, the error is upstream_line_unavailable. Failed tasks are not charged and reserved credits are refunded; retry later. This error does not identify a specific parameter or material problem. Explicit duration limits, unreadable references and content-review rejections retain duration_rejected, reference_unreadable and content_policy respectively; follow the relevant guidance.

## FAQ

### What is KovAPI?

KovAPI is an OpenAI-compatible AI API platform: one key gives you chat, image and video models. Existing OpenAI SDK code connects by changing the base URL to https://kovapi.com/v1 and using your KovAPI key, and Claude models also speak the native Anthropic API.

### How are NezhaGate and KovAPI related?

KovAPI is the new name of NezhaGate, now at kovapi.com. Your account, balance, API keys and prices are unchanged, and the nezhagate.com API address keeps working, so existing code needs no changes. On the website, sign in once at kovapi.com.

### What does "chat, images, video — one API" actually mean?

One Base URL and one API key cover all three kinds of work: chat at /v1/chat/completions, images at /v1/images/generations (image-to-image at /v1/images/edits), and video at /v1/videos/generations. The models come from OpenAI (GPT), Anthropic (Claude), Google (Gemini), ByteDance (Seedance), Alibaba (Wan), xAI (Grok), MiniMax and more, so you do not integrate an SDK, open an account and reconcile a bill for each vendor separately. To switch or compare models, you change a single model parameter.

### How does KovAPI differ from the official OpenAI / Claude / Gemini APIs?

KovAPI serves the vendors' own models; what changes is how you reach them. One key covers every vendor, the API is OpenAI-compatible across the board (Claude also speaks the native Anthropic API), and billing, logs and balance live in one console. Flagship models are connected to multiple suppliers with automatic failover, and you save up to 85% against official prices.

### Which models are supported?

KovAPI currently offers OpenAI's GPT-6 family (Astra / 6.1 Sol / Sol / Luna), the GPT-5.6 family (Sol / Luna) and the image models GPT Image 2 and GPT Image 2.5 (Flare / Sunburst); Anthropic's Claude Opus 5.5, Claude Sonnet 5.5, Claude Opus 5, Claude Sonnet 4.6 and Claude Fable 5; Google's Gemini 3.1 Pro, Gemini 3.8 Flash and the image models Nano Banana 2.1, Nano Banana 2 and Nano Banana Pro; DeepSeek V4.1 Flash and V4 Flash 0731; Z.ai's GLM-5.3; Alibaba's Qwen3.8 family (Max / Flash); xAI's Grok 4.7 and the image model Grok Imagine Image (Standard / Quality); and for video, Seedance 2.5 and the Seedance 2.0 family, Wan 3.0, MiniMax H3 and Grok Imagine Video 1.5. Every model shares the same endpoint and key; the live, complete list is on the Model Marketplace.

### Are KovAPI's models the original vendor models?

Yes. The model you ask for is the model that answers, and we never rewrite its output. The one exception is the Chinese models GLM and Qwen: if one of them cannot answer for a while, the request is answered by a model of equal or greater capability, billed at whichever of the two prices is lower, and your call log says which model answered and which price applied. Every call is in your logs with the model, usage, charge and latency, and each model's live state is public on the Status page. Every week we also run a fixed question set against our flagship models to measure success rate, time to first token and output speed.

### Do you support GPT, Claude and Gemini?

Yes. The main models from GPT (OpenAI), Claude (Anthropic) and Gemini (Google) are all available through one OpenAI-compatible interface. Switching models only requires changing the model field in your request — no new SDK or re-integration needed.

### Do you support image and video generation?

Yes. For images there are GPT Image 2, GPT Image 2.5 (Flare / Sunburst), Nano Banana 2.1, Nano Banana 2 and Nano Banana Pro: text-to-image at /v1/images/generations and image-to-image at /v1/images/edits, up to 4K. For video there are Seedance, Wan 3.0, MiniMax H3 and Grok Imagine Video at /v1/videos/generations. Results come back as direct links you can use in your app right away.

### How often are prices and model details updated?

The Pricing page and the Model Marketplace are wired to the billing system in real time, so the price you see is exactly the price you are charged. New models and price changes appear as soon as they go live, and price changes are announced on the site one day in advance.

### How is KovAPI priced? Are there hidden fees?

KovAPI bills by actual usage: chat by the token, images by the picture, and video mostly by the second (a few models are priced per clip, as the Pricing page shows). You save up to 85% against official prices, and every model is listed next to its official price on the Pricing page — no hidden multipliers, no prices that only appear after you top up. Price changes are announced on the site one day in advance.

### Why is KovAPI cheaper than the official APIs?

KovAPI buys model capacity in bulk from multiple suppliers at below-retail cost, so most models cost less here than at the official API; the Pricing page shows exactly how much you save on each one. We earn the margin between our cost and our price, and we charge no top-up fees.

### How are failed calls billed?

KovAPI never charges for a failed call. Asynchronous jobs such as images and video reserve credits when you submit them and refund them in full automatically if the job fails, with the reason written to your log. If a streamed answer breaks off midway, you are billed only for the usage the upstream actually returned.

### Do KovAPI credits expire? Can I get a refund?

KovAPI credits don't expire. Top-ups start at ¥10 (260 credits) with WeChat Pay and at $5 with USDT or a card, and every new account gets 30 free credits at sign-up and 50 more with its first top-up. Your top-up history and the charge for every call in the last 60 days are in the console. Within 7 days of a top-up, unused purchased credits can be refunded through support, minus a fee per refund of $1.40 for top-ups made on or after October 29, 2026 and $0.40 for earlier ones; bonus credits are not refundable, and USDT payments can be refunded too.

### How do I check my balance and usage?

Sign in and the console shows your live balance, usage per API key, and every call from the last 60 days: model, time, tokens or image count, latency and charge, with the reason written next to any failed call. You can also read your balance programmatically through the balance endpoint.

### How reliable is KovAPI?

KovAPI connects flagship models — GPT, Gemini, DeepSeek, GLM, Kimi, GPT Image, Nano Banana, Seedance 2.0 and more — to multiple suppliers. If one returns an error, times out or is at capacity, the request moves automatically to the next supplier for the same model; a streamed answer switches before it starts and is never moved once output has begun. The gateway is probed around the clock, every 2 minutes; live model status and gateway availability are published on the Status page, with success rates counted from real customer traffic only. Failed calls are never billed, and failed image and video jobs are refunded in full.

### Does KovAPI have rate limits?

KovAPI sets no default per-account rate limit. On the API Keys page you can give each key its own requests-per-minute cap, daily budget, total spend limit, model allowlist and IP allowlist. If a model is at capacity, you get a 429 with a Retry-After header; retry after that interval. For sustained high concurrency, contact support in advance and we will add capacity for your volume. Free promotional models carry a daily call quota.

### Is it suitable for commercial projects or AI tool sites?

Yes. KovAPI gives you an OpenAI-compatible API, per-project API keys (each with its own spend, rate, model and IP limits), itemised call logs and billing that never charges for failures, ready to plug into commercial products and tool sites. Please follow the Terms of Service and each model vendor's usage policies.

### Do you support enterprise cooperation or bulk calls?

Yes. KovAPI can handle larger-scale batch calls, and we welcome enterprise inquiries. For higher concurrency, dedicated quota or integration needs, email support@kovapi.com or use the Contact us option on the site. The plan is tailored to your usage and scenario.

### Which scenarios suit KovAPI best?

KovAPI suits cases where you need to call several model providers in one project, want unified billing and usage management, or want to quickly validate different models. Common uses include AI tool sites, SaaS products, automation workflows, agents and multi-model comparison. It is especially convenient when you need to switch flexibly between models.

### How do I get started quickly with KovAPI?

Getting started takes three steps: register an account, create an API key in the console, then point your Base URL at KovAPI and add the key. Because it is fully OpenAI-compatible, existing OpenAI SDK code needs almost no changes — just swap base_url and api_key. See the developer docs for details.

### How should I configure the Base URL?

Set the Base URL to https://kovapi.com/v1. In the OpenAI SDK, set base_url (or the OPENAI_BASE_URL environment variable) to that address and use the key you created in the console. Use /v1/chat/completions for chat and /v1/images/generations for images — the paths match OpenAI.

### Is it compatible with the OpenAI SDK?

Yes. KovAPI implements the OpenAI-compatible interface, so the official openai SDK works by simply changing base_url and api_key. Request and response shapes match OpenAI, including streaming output and multimodal input.

### Do you support Python, Node.js and curl?

Yes. Any language that can make HTTP requests or use the OpenAI SDK works, including Python, Node.js, Go, Java and plain curl. The docs include ready-to-run Python, Node.js and curl examples.

### Do you support n8n, Dify, LangChain, Open WebUI, LobeChat, Cherry Studio and ChatBox?

Yes. Most of these tools let you set a custom OpenAI-compatible Base URL and API key — point them at KovAPI to use the platform's models. If a tool supports an OpenAI-compatible or custom endpoint, it can connect to KovAPI with no extra plugin.

### What if my API key is leaked?

If you suspect an API key is leaked, delete or rotate it in the console immediately — once revoked, the old key can no longer be used. We recommend creating separate keys per project and never embedding keys in front-end code or public repos. You can check each key's usage in the console at any time to spot anomalies.

### Does KovAPI store or look at my data?

KovAPI uses call content only to complete the call, to show it to you in your logs and to troubleshoot problems; it is never used to train any model and never sold or shared with advertisers. Apart from the model suppliers and infrastructure services needed to complete a call (CDN, cloud storage, and a search service when you turn on web search), it is not shared with third parties. Call logs stay in your console for 60 days; their text content is deleted after 30 days, leaving only the usage record needed for billing. Generated images and videos are deleted after 15 days and uploaded reference files after 7 days. Database backups are kept for up to 30 more days.

### Are my account and API keys safe?

Every API key can carry its own spend limit, requests-per-minute cap, model allowlist and IP allowlist, and can be disabled or deleted at any time, so one leaked key never reaches your other projects. Keys are stored encrypted on our servers and you can view and copy them whenever you are signed in; every call and every balance change is itemised for you to check.

### Who do I contact if something goes wrong?

KovAPI support is under "Contact" in the site's top navigation and in your console: WeChat, our Telegram group, and email at support@kovapi.com. Support hours are 10:00–19:00 Beijing time (UTC+8), and we reply within an hour of seeing your message.

### What uses are prohibited?

You may not use KovAPI for anything that violates laws or the model vendors' usage policies, including generating or distributing illegal, infringing, fraudulent or malware content, or automated abuse and bypassing quota or safety limits. Violations may lead to account restriction or termination. The full rules are in the Terms of Service.

### What should I keep in mind when using KovAPI?

Keep your API keys safe, top up as you need, and follow the Terms of Service and each model vendor's usage policies. Check unit prices on the Pricing page before you call, and give your keys a daily budget and a total limit to rule out unexpected spend.

## Optional
- https://kovapi.com/en/llms.txt
- https://kovapi.com/en/pricing
- https://docs.kovapi.com/en/guide
- https://kovapi.com/en/faq