跳至正文
OpenAgentSkill
浏览已开放 · 目录记录日期 2026-11-21 · 自动更新和账户服务暂时关闭。
浏览器与自动化

venice-models

Discover Venice models, their capabilities, constraints, and pricing. Covers GET /models (with ?type filter), /models/traits, /models/compatibility_mapping, the ModelResponse schema (capabilities, constraints, pricing per type), and how to use this to pick the right model programmatically.

veniceai · 已有 AI 审核记录

安装与使用

运行命令前先阅读来源与依赖说明。复制命令不会自动安装或执行。

npx skills add veniceai/skills --skill venice-models

技能说明


name: venice-models description: Discover Venice models, their capabilities, constraints, and pricing. Covers GET /models (with ?type filter), /models/traits, /models/compatibility_mapping, the ModelResponse schema (capabilities, constraints, pricing per type), and how to use this to pick the right model programmatically.

Venice Models

Three read-only endpoints for model discovery — all GET:

EndpointReturns
/modelsFull model catalog with model_spec (capabilities, constraints, pricing).
/models/traitsTrait → model ID mapping (e.g. "default", "fastest", "default_reasoning", "highest_quality").
/models/compatibility_mappingLegacy / OpenAI / third-party model ID → Venice model ID aliases.

All three take an optional ?type= filter: text, image, video, music, tts, asr, embedding, upscale, inpaint, all, code.

All three are authenticated (Bearer API key or x402 SIWE) like every other /api/v1 route.

Use when

  • You need to pick a model at runtime based on capabilities (vision, reasoning, function calling, E2EE, X search, multi-image, …).
  • You need to validate a request against a model's constraints (prompt length, aspect ratio, resolution, steps).
  • You need the current price per million tokens / per image / per second / per 1k chars to build a cost estimate.
  • You want to resolve a user-friendly trait name (e.g. default, default_reasoning, highest_quality) or a frontier-style ID (openai-gpt-54-pro, claude-opus-4-7) to a concrete Venice model ID.

GET /models

curl "https://api.venice.ai/api/v1/models?type=text"
{
  "object": "list",
  "type": "text",
  "data": [
    {
      "id": "zai-org-glm-5-1",
      "created": 1699000000,
      "model_spec": {
        "name": "GLM 5.1",
        "description": "Balanced blend of speed and capability...",
        "availableContextTokens": 200000,
        "maxCompletionTokens": 24000,
        "privacy": "private",
        "beta": false,
        "betaModel": false,
        "modelSource": "https://huggingface.co/zai-org/GLM-5.1",
        "offline": false,
        "capabilities": { ... },
        "constraints": { ... },
        "pricing": { ... },
        "regionRestrictions": ["US"],
        "deprecation": {"date": "2025-03-01T00:00:00.000Z"}
      }
    }
  ]
}

model_spec.capabilities — text models

FlagMeaning
optimizedForCodeTuned for coding tasks.
quantizationfp4 / fp8 / fp16 / bf16 / int8 / int4 / not-available.
supportsFunctionCallingTools are allowed.
supportsReasoningEmits <thinking>...</thinking> blocks (and/or provider-specific reasoning_content).
supportsReasoningEffortHonors reasoning.effort / reasoning_effort.
supportsResponseSchemaHonors response_format: json_schema.
supportsMultipleImages + maxImagesMulti-image vision support.
supportsVisionAccepts image_url parts.
supportsVideoInputAccepts video_url parts.
supportsWebSearchHonors venice_parameters.enable_web_search.
supportsLogProbsHonors logprobs / top_logprobs.
supportsTeeAttestationRuns inside a TEE with hardware attestation.
supportsE2EEEnd-to-end encrypted inference available (requires TEE).
supportsXSearchxAI native web + X/Twitter search.
supportsAudioInputAccepts audio-content message parts (set by the runtime capability builder — not part of the OpenAPI strict schema but appears on /models responses).

model_spec.constraints — by model family

  • Text — temperature.default, top_p.default, and {frequency,presence,repetition}_penalty.default.
  • Image — promptCharacterLimit, widthHeightDivisor, steps.{default,max}, optional aspectRatios[] + defaultAspectRatio, optional resolutions[] + defaultResolution (the last two appear only on models that use ratio/resolution-based sizing).
  • Video — aspect_ratios[], resolutions[], durations[], model_type (text-to-video/image-to-video/video), audio, audio_configurable, prompt_character_limit.
  • Inpaint / edit — aspectRatios[], promptCharacterLimit, combineImages.
  • TTS / Music (fields surface at the top level of model_spec, not inside constraints) — voices[], default_voice, supports_lyrics, lyrics_required, supports_lyrics_optimizer, supports_force_instrumental, supports_speed, supports_language_code, min_speed, max_speed, min_prompt_length, prompt_character_limit. Internal TTS per-model toggles like supportsPromptParam / supportsTemperatureParam / supportsTopPParam exist on the model definitions but are not merged into /models output today — treat the speech request schema as the support matrix.
  • Embedding (top-level, not inside constraints) — embeddingDimensions, maxInputTokens, supportsCustomDimensions.

model_spec.pricing — by model family

  • LLM — input.{usd,diem}, output.{usd,diem} per 1 000 000 tokens, plus optional cache_input (reads), cache_write (writes, e.g. Anthropic 1.25×), and extended.* tier triggered by context_token_threshold.
  • Image — either generation.{usd,diem} per image (flat) or resolutions.<tier>.{usd,diem} (per 1K/2K/4K). Every image row also carries a global upscale.{2x,4x}.{usd,diem} block (derived from shared upscale SKUs) — treat it as account-wide upscale pricing, not a signal that this specific model can upscale. Combine with the inpaint/upscale model's own capability check to decide what's actually callable.
  • Inpaint / edit — inpaint.{usd,diem} per edit.
  • Video — not currently returned on /models. calculatePricing() has no video branch, so video entries have no model_spec.pricing. Use POST /video/quote for the authoritative per-request price.
  • Music / long audio — generation.{usd,diem} (per job)
阅读完整来源 ↗