# EvoLink

> EvoLink is an AI model gateway: one API key for 170+ models — LLMs (GPT, Claude, Gemini, GLM, DeepSeek) plus image, video and music generation — automatically routed to the lowest available price. Endpoints are OpenAI-, Anthropic- and Gemini-compatible, so agents and existing SDKs connect by changing only the base URL and key.

## Workflow examples

- [Claude Fable 5.5 use cases and adapted prompts](https://evolink.ai/claude-fable-5-5-prompts): source-linked community demos with execution-environment requirements and EvoLink-adapted evaluation prompts. Model attribution is not independently verified; official Fable 5.5 release and EvoLink access remain unconfirmed.

## Connect an agent

- Base URLs: Anthropic-style clients use https://direct.evolink.ai · OpenAI-compatible clients use https://direct.evolink.ai/v1 · Gemini-style clients use https://direct.evolink.ai/v1beta
- Auth: Bearer API key ("Authorization: Bearer <key>"; Anthropic-style also accepts x-api-key). The hosted MCP server and the EvoLink CLI sign in through the browser (OAuth) instead; only the local stdio MCP packages and scene skills read the EVOLINK_API_KEY environment variable.
- [Create an API key](https://evolink.ai/signup?returnUrl=/dashboard/keys&utm_source=llms_txt): takes about 30 seconds.

## Machine-readable model catalog

- Kimi K3.1 Pro and Highspeed: use kimi-k3.1-pro or kimi-k3.1-pro-highspeed through /v1/chat/completions, /v1/responses or /v1/messages, with JSON or SSE. Both offer 1,048,576-token context, text and visual input, tool calling, always-on reasoning with low/high/max effort, 5m/1h prompt-cache writes and separately billed built-in web search. Highspeed has higher input/output/cache-write rates; cache reads and search share the same rates. Compare measured latency and cost for your workload. [Model page and calculator](https://evolink.ai/kimi-k3-1-pro) · [API reference and search examples](https://evolink.ai/docs/en/api-manual/language-series/kimi/chat-completions/chat-completions-quickstart)
- [Model list, OpenAI format](https://direct.evolink.ai/v1/models): GET with Bearer key; returns model ids with supported_endpoint_types.
- MiniMax H3 Max Turbo: text-to-video and image-to-video at 480p/768p for integer durations of 5-15 seconds; image-to-video requires image_start and accepts optional image_end. Use minimax-h3-max-turbo-text-to-video or minimax-h3-max-turbo-image-to-video with POST https://api.evolink.ai/v1/videos/generations. Per-output-second pricing; no reference media or 2K output. [Model page & current pricing](https://evolink.ai/minimax-h3-max-turbo) · [API specification](https://evolink.ai/api-reference/minimax-h3-max-turbo/minimax-h3-max-turbo-text-to-video.md) · [Hailuo family comparison](https://evolink.ai/collections/hailuo)
- GLM-5.3 family: use glm-5.3-flash for high-volume and multimodal work (text, image, video, files), glm-5.3-flashx when the same multimodal input needs more throughput at about a quarter of the flagship price, and glm-5.3 for difficult text-only coding and long-horizon engineering; all three expose 1M context, always-on reasoning, tool calling and prompt caching through the unified API. [GLM-5.3 details](https://evolink.ai/glm-5-3) · [GLM-5.3 Flash details](https://evolink.ai/glm-5-3-flash) · [GLM-5.3 FlashX details](https://evolink.ai/glm-5-3-flashx) · [Family routing guide](https://evolink.ai/blog/glm-5-3-flash-vs-glm-5-3)
- Featured LLM: claude-fable-5 (Claude Fable 5, the previous Fable generation) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-fable-5". [Model details](https://evolink.ai/claude-fable-5) · [Launch page](https://evolink.ai/launch/fable-5)
- Newest Anthropic flagship: claude-fable-5-1 (Claude Fable 5.1 — Anthropic's most capable model, the Fable tier above Opus; 1M context, 128K max output, prompt caching read + write with cache reads at a quarter of Fable 5's rate, web search tool; 5% below official token pricing) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-fable-5-1"; Anthropic-style: POST https://direct.evolink.ai/v1/messages. [Model details & pricing](https://evolink.ai/claude-fable-5-1)
- Newest Anthropic Opus: claude-opus-5-5 (Claude Opus 5.5 — succeeds Claude Opus 5 at a lower list price; 1M context, 128K max output, always-on thinking controlled by effort (default medium), 5-minute cache writes and separately billed cache reads, web search tool; input, output, and 5-minute cache writes are 5% below official pricing; cache reads use the displayed rounded rate) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-opus-5-5"; Anthropic-style: POST https://direct.evolink.ai/v1/messages. [Model details & pricing](https://evolink.ai/claude-opus-5-5)
- Newest Anthropic Sonnet: claude-sonnet-5-5 (Claude Sonnet 5.5 — succeeds Claude Sonnet 5 at the same official list price; 1M context, 128K max output, adaptive thinking on by default with between_tools as the lowest setting (thinking: disabled is not accepted), recalibrated effort levels (default high), 5-minute cache writes and separately billed cache reads; web search depends on the route, so confirm support before sending web_search; input, output, and 5-minute cache writes are 5% below official pricing; cache reads use the displayed rounded rate) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-sonnet-5-5"; Anthropic-style: POST https://direct.evolink.ai/v1/messages. [Model details & pricing](https://evolink.ai/claude-sonnet-5-5)
- Newest Anthropic Haiku: claude-haiku-5-5 (Claude Haiku 5.5 — succeeds Claude Haiku 4.5; 1M context, 128K max output, adaptive thinking on by default with effort as the control (default medium; thinking: disabled is accepted at effort high or below), two rate cards by prompt length (base rates up to 100K prompt tokens, 5x above, counting input plus cache writes and cache reads), 5-minute cache writes and separately billed cache reads; thinking budgets and non-default sampling values are not accepted; web search depends on the route, so confirm support before sending web_search; input, output, and 5-minute cache writes are priced slightly below official list rates; cache reads use the displayed rounded rate) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-haiku-5-5"; Anthropic-style: POST https://direct.evolink.ai/v1/messages. [Model details & pricing](https://evolink.ai/claude-haiku-5-5)
- Previous Opus flagship: claude-opus-5 (Claude Opus 5 — 1M context, 128K max output, prompt caching read + write, web search tool; 5% below official token pricing) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "claude-opus-5"; Anthropic-style: POST https://direct.evolink.ai/v1/messages. [Model details & pricing](https://evolink.ai/claude-opus-5)
- Newest Google Gemini model: gemini-4-argon (Gemini 4 Argon — announced by Google on 2026-09-30; 1,048,576-token context window shared by input, thinking, and output; a single request returns up to 65,536 thinking and output tokens (32,768 without streaming), and outputs up to 1,048,576 tokens continue across requests on the Gemini-native API with continuationToken; cached input billed at a lower rate than input; Google Search billed per search query) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "gemini-4-argon" · Gemini-native: POST https://direct.evolink.ai/v1beta/models/gemini-4-argon:generateContent. [Model page & current pricing](https://evolink.ai/gemini-4) · [Gemini family comparison](https://evolink.ai/collections/gemini)
- Newest Google Flash LLM: gemini-3.8-flash (Gemini 3.8 Flash — GA 2026-09-02; 1,048,576-token context, 65,536 max output; text, image, video, audio and PDF input; thinking levels low / medium / high, prompt caching, function calling, structured output) — OpenAI-compatible: POST https://direct.evolink.ai/v1/chat/completions with model "gemini-3.8-flash" · Gemini-native: POST https://direct.evolink.ai/v1beta/models/gemini-3.8-flash:generateContent. [Model page & current pricing](https://evolink.ai/gemini-3-8-flash) · [3.8 vs 3.7 Flash decision guide](https://evolink.ai/blog/gemini-3-8-flash-vs-gemini-3-7-flash) · [API integration guide](https://evolink.ai/blog/how-to-use-gemini-3-8-flash-api) · [Gemini family comparison](https://evolink.ai/collections/gemini)
- Newest video flagship: Seedance 2.5 is live on EvoLink (five routes plus a draft-to-1080p conversion; 1080p verified 2026-08-17) — 4-30s one-shot clips, 480p/720p/1080p, synchronized audio, up to 50 reference inputs across six model IDs (seedance-2.5-text-to-video / image-to-video / reference-to-video / video-edit / video-extend, plus seedance-2.5-draft-to-video to upgrade a completed 480p draft to 1080p); async task flow via POST https://direct.evolink.ai/v1/videos/generations. [Model page & pricing](https://evolink.ai/seedance-2-5) · [Integration docs](https://evolink.ai/docs/en/api-manual/video-series/seedance2.5/seedance-2.5-text-to-video) · [2.5 vs 2.0 decision guide](https://evolink.ai/blog/seedance-2-5-vs-seedance-2-0) · [120+ copy-ready Seedance 2.5 prompts with video examples](https://evolink.ai/seedance-2-5-prompts)
- Newest 4K video route: Gemini Omni 1.2 Flash is live on EvoLink — 3-15s clips at 360p/720p/1080p/4K with synchronized audio, first-and-last-frame image control, reference videos, video edit and extend, plus standalone upscaling of 360p/720p/1080p clips up to 4K, across six model IDs (gemini-omni-1.2-flash-text-to-video / image-to-video / reference-to-video / video-edit / video-extend / video-upscale); token-based billing, async task flow via POST https://direct.evolink.ai/v1/videos/generations. Note: video-upscale takes no prompt and needs an exact 640x360, 1280x720 or 1920x1080 input. [Model page & pricing](https://evolink.ai/gemini-omni-1-2-flash)
- Previous Omni route: Gemini Omni 1.1 Flash is live on EvoLink (five routes) — 3-10s clips at native 360p/720p or upscaled 1080p/4K, with synchronized audio, first-and-last-frame image control, up to three reference videos, plus video edit and extend across five model IDs (gemini-omni-1.1-flash-text-to-video / image-to-video / reference-to-video / video-edit / video-extend); token-based billing, async task flow via POST https://direct.evolink.ai/v1/videos/generations. Note: for video-extend, duration is the number of seconds APPENDED, not the total length. [Model page & pricing](https://evolink.ai/gemini-omni-1-1-flash)
- Featured image model: doubao-seedream-5.0-pro (Seedream 5.0 Pro — 1K/2K output tiers, up to 10 reference images, instruction-based editing; input images billed per image) — POST https://direct.evolink.ai/v1/images/generations. [Model page & pricing](https://evolink.ai/seedream-5-0-pro) · [Multi-reference workflow guide](https://evolink.ai/blog/how-to-use-seedream-5-0-pro-multi-reference-image-generation)
- New image model: mj-v8.2 (Midjourney V8.2 — 4 images per request, Draft/Fast speed, Standard/HD quality, free --q detail 1-4 in the prompt; sibling model IDs mj-v8.2-variation / -remix / -edit / -retexture / -upload-paint / -remove-bg; per-request billing, failed tasks refunded) — POST https://direct.evolink.ai/v1/images/generations. [Model page & pricing](https://evolink.ai/midjourney-v8-2)
- New image model: doubao-seedream-5.0-flash (Seedream 5.0 Flash — the speed-focused Seedream 5.0 generation and editing route (standard prompt mode only, one image per request); one flat price per image at 1K/1.5K/2K, up to 10 reference images that are not billed; layer split via doubao-seedream-5.0-flash-layerize, billed per output image) — POST https://direct.evolink.ai/v1/images/generations. [Model page & pricing](https://evolink.ai/seedream-5-0-flash)
- New image model: grok-imagine-image-2.0 (xAI Grok Imagine Image 2.0 — text-to-image and 1-3 reference-image editing in one model, Low/Medium quality, 1K/2K resolution, 1-10 images per request) — POST https://direct.evolink.ai/v1/images/generations. [Model page & pricing](https://evolink.ai/grok-imagine-image-2-0)
- [Pricing catalog, anonymous](https://api.evolink.ai/web/api/models/pricing): GET, no auth; returns the whole catalog in one JSON response — no pagination and no query filters — as a "data" array of SKUs with model_name, model_type (text/image/video/audio), billing_rule, prices (USD/CNY/credits) and metadata. For one model only: GET https://api.evolink.ai/web/api/models/pricing/{model_name} (e.g. .../pricing/claude-opus-5).
- [Human-readable model pages](https://evolink.ai/models)

## MCP and CLI for media generation (official)

- [MCP landing page](https://evolink.ai/mcp): the current way to give an agent EvoLink's image, video and audio models — pick the client, copy its setup prompt, sign in once in the browser. Agent-readable summary: https://evolink.ai/mcp/agent-overview/en.md
- Remote MCP server: https://mcp.evolink.ai/mcp (Streamable HTTP; browser OAuth sign-in, no API key; 16 tools — search_models / recommend_models, get_model, estimate_cost, generate_image / generate_video / generate_audio, get_task, list_tasks, check_balance, upload_file, prepare_upload, get_upload and more). Setup by client (Claude, ChatGPT, Codex, Claude Code, Cursor, OpenClaw, Hermes): [MCP guide](https://evolink.ai/docs/en/mcp/overview) · [tool reference](https://evolink.ai/docs/en/mcp/tools)
- EvoLink CLI + skill for agents that run terminal commands (Codex, Claude Code, Cursor, OpenClaw, Hermes, Gemini CLI, OpenCode, Copilot CLI), SSH sessions and scripts: npm install -g @evolinkai/cli · evolink auth login · evolink skills install (Node.js 22+; browser sign-in, no API key; the bundled evolink-cli skill teaches the agent the commands). [CLI setup](https://evolink.ai/docs/en/cli/setup) · [command reference](https://evolink.ai/docs/en/cli/reference)
- Which route: browser chat (Claude, ChatGPT) and native-MCP hosts → remote MCP; agents with a terminal → CLI; use one route per task. [Decision table](https://evolink.ai/docs/en/cli-mcp/overview)
- Local stdio MCP, only for clients without remote-HTTP/OAuth support: npx -y @evolinkai/mcp@latest with EVOLINK_API_KEY (Node.js 18+; 14 tools, no one-time upload tools; local file_path uploads need EVOLINK_UPLOAD_ALLOWED_DIRS). [Local MCP setup](https://evolink.ai/docs/en/mcp/other-clients#local-mcp)
- Text-model routing MCP (API key, separate from media): [@evolinkai/mcp-router](https://www.npmjs.com/package/@evolinkai/mcp-router) — npx -y @evolinkai/mcp-router@latest; smart_route (read-only), delegate, cascade.

## Agent skills and plugins

- CLI skill evolink-cli: installed by evolink skills install into the host's skill directory (~/.agents/skills, ~/.claude/skills, ~/.openclaw/skills, ~/.hermes/skills). [CLI skills](https://evolink.ai/docs/en/cli/skills)
- MCP plugin and skill evolink-mcp: Codex "codex plugin marketplace add Evolink-AI/mcp-plugin" then "codex plugin add evolink@evolink"; Claude Code "/plugin marketplace add Evolink-AI/mcp-plugin" then "/plugin install evolink@evolink"; skill only: npx skills add Evolink-AI/mcp-plugin -g -y. [Plugin guide](https://evolink.ai/docs/en/mcp/plugins) · [source](https://github.com/Evolink-AI/mcp-plugin)
- Shared agent rules, Markdown and llms.txt resources: [agent resources](https://evolink.ai/docs/en/cli-mcp/agent-resources)
- Scene skills for ad and product visuals (API key; GPT Image 2 / Seedance 2 workflows): [EvoLinkAI/evolink-agent-skills](https://github.com/EvoLinkAI/evolink-agent-skills) — npx skills add EvoLinkAI/evolink-agent-skills --skill <name> -g

## Docs

- [CLI / MCP overview](https://evolink.ai/docs/en/cli-mcp/overview): choose a route, install by client, verify for free, references and troubleshooting
- [Developer documentation](https://evolink.ai/docs) — full API reference and integration guides ([docs llms.txt](https://evolink.ai/docs/llms.txt))
- [LLM API integration guides](https://evolink.ai/docs/en/integration-guide/claude-code-cli): use EvoLink as the model provider in Claude Code CLI, Codex CLI, Gemini CLI, OpenClaw, OpenCode, Hermes, Claude Desktop, CodeBuddy

## New models

- Grok Imagine Video 1.5 Lite: xAI video generation from text or one first-frame image, with native audio and 1–15 second clips up to 1080p; reference images are not supported. [Model page & pricing](https://evolink.ai/grok-imagine-video-1-5-lite) · [API reference](https://evolink.ai/api-reference/grok-imagine-video-1.5-lite/grok-imagine-video-1.5-lite-API-Reference.md)
- GPT-6 Luna Decisions: OpenAI's Decisions API (POST /v1/decisions) on GPT-6 Luna, live on EvoLink (model ID gpt-6-luna-decisions). Typed answers (probability, choice, or score) for text and inline images in one synchronous call; billed on input tokens only. [Model page & pricing](https://evolink.ai/gpt-6-luna-decisions) · [API reference](https://evolink.ai/docs/en/api-manual/language-series/gpt/decisions/decisions-reference)
- GPT-6.1 Terra: released by OpenAI on October 26, 2026 (model ID gpt-6.1-terra). Visual reasoning, computer use and writing; 1.05M shared context, 128K max output, reasoning effort none to max, Responses for tools and Chat Completions without tools. [Model page & pricing](https://evolink.ai/gpt-6-1-terra)
- GPT-6.1 Sol: released by OpenAI on September 29, 2026 (model ID gpt-6.1-sol). Multi-file coding and tool-driven agents; 1.05M shared context, 128K max output, Responses for tools and Chat Completions without tools. [Configured model page & pricing](https://evolink.ai/gpt-6-1-sol) · [Release and channel status](https://evolink.ai/blog/gpt-6-1-sol-release-date) · [GPT-6.1 Sol vs GPT-6 Sol](https://evolink.ai/blog/gpt-6-1-sol-vs-gpt-6-sol)
- GPT-6 Sol: released by OpenAI on September 22, 2026 and live on EvoLink (model ID gpt-6-sol, 1.05M context, complex coding and agents). [Model page & pricing](https://evolink.ai/gpt-6-sol) · [Release date & rollout](https://evolink.ai/blog/gpt-6-sol-release-date) · [GPT-6 Sol vs GPT-5.6 Sol](https://evolink.ai/blog/gpt-6-sol-vs-gpt-5-6-sol)
- GPT-6 Luna: released by OpenAI on September 22, 2026 and live on EvoLink (model ID gpt-6-luna, 1.05M context, high-volume tasks). [Model page & pricing](https://evolink.ai/gpt-6-luna) · [Release date & rollout](https://evolink.ai/blog/gpt-6-luna-release-date) · [GPT-6 Luna vs GPT-5.6 Luna](https://evolink.ai/blog/gpt-6-luna-vs-gpt-5-6-luna)

## Optional

- [Full overview in one document](https://evolink.ai/llms-full.txt): billing model, async video task flow, model specs and prices without crawling individual pages.
- [Pricing overview](https://evolink.ai/models)
- [Blog](https://evolink.ai/blog)
- [API changelog](https://evolink.ai/changelog)