AI Models

LenserFight routes lens executions through a unified model registry (ai.models). Each model has a canonical key used in API calls, a set of capabilities, and declared input/output modalities.

Capability tags

Tag	Meaning
chat	Conversational / instruction-following text generation
reasoning	Extended chain-of-thought reasoning (think-before-answer)
tools	Function / tool calling
vision	Accepts image inputs
json_schema	Provider-supported structured JSON output
image_generation	Produces images from text prompts
video_generation	Produces video clips from text/image prompts
audio_generation	Produces speech or general audio
music_generation	Produces music or soundtracks

---

OpenAI

GPT-5.4 Pro

Field	Value
Key	gpt-5.4-pro
Capabilities	chat · reasoning · tools · vision · json_schema
Context window	400 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Most capable GPT-5.4 tier. Use for complex multi-step tasks, agentic workflows, and vision-heavy lens designs.

GPT-5.4

Field	Value
Key	gpt-5.4
Capabilities	chat · reasoning · tools · vision · json_schema
Context window	400 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

GPT-5.4 Mini

Field	Value
Key	gpt-5.4-mini
Capabilities	chat · tools · json_schema
Context window	400 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Faster, cost-efficient variant. No vision or reasoning.

GPT-5.4 Nano

Field	Value
Key	gpt-5.4-nano
Capabilities	chat · json_schema
Context window	400 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Lowest-latency OpenAI text model. Best for simple classification or extraction pipelines.

GPT-5.2

Field	Value
Key	gpt-5.2
Capabilities	chat · reasoning · tools · vision · json_schema
Context window	400 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

GPT-4o

Field	Value
Key	gpt-4o
Capabilities	chat · tools · vision · json_schema
Context window	128 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

---

OpenAI — Generative Media

DALL-E 4

Field	Value
Key	dall-e-4
Capabilities	image_generation
Input modalities	text
Output modalities	image
Provider docs	—

Synchronous image generation. Returns a signed URL immediately after the API call resolves.

Sora 2.0

Field	Value
Key	sora-2.0
Capabilities	video_generation
Input modalities	text
Output modalities	video
Provider docs	—

Async video generation. The execution engine returns a pending task ID; the lenser polls until the clip is ready.

---

Anthropic

Claude Opus 4.6

Field	Value
Key	claude-opus-4-6
Capabilities	chat · reasoning · tools
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Anthropic's most capable model. Suited for long-document analysis, multi-step reasoning chains, and complex agentic tasks.

Claude Sonnet 4.6

Field	Value
Key	claude-sonnet-4-6
Capabilities	chat · reasoning · tools
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Balanced speed/quality. Default recommendation for most Claude-backed lenses.

Claude Sonnet 4.5

Field	Value
Key	claude-sonnet-4-5
Capabilities	chat · reasoning · tools
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Claude Sonnet 4.0

Field	Value
Key	claude-sonnet-4-0
Capabilities	chat · reasoning · tools
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Claude Haiku 4.5

Field	Value
Key	claude-haiku-4-5
Capabilities	chat
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Fastest, cheapest Anthropic model. Use for high-volume classification or summarisation where speed matters most.

Claude Haiku 3.5

Field	Value
Key	claude-haiku-3-5
Capabilities	chat
Context window	200 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Legacy. Prefer Haiku 4.5 for new lenses.

---

Google

Gemini 3.1 Pro Preview

Field	Value
Key	gemini-3.1-pro-preview
Capabilities	chat · reasoning · tools · vision
Context window	2 000 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Gemini 3 Pro Preview

Field	Value
Key	gemini-3-pro-preview
Capabilities	chat · reasoning · tools · vision
Context window	2 000 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Gemini 2.5 Pro

Field	Value
Key	gemini-2.5-pro
Capabilities	chat · reasoning · tools · vision
Context window	2 000 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Industry-leading 2M-token context. Excellent for large codebase analysis or full-book summarisation.

Gemini 3 Flash Preview

Field	Value
Key	gemini-3-flash-preview
Capabilities	chat · tools · vision
Context window	1 000 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Gemini 2.5 Flash

Field	Value
Key	gemini-2.5-flash
Capabilities	chat · tools · vision
Context window	1 000 000 tokens
Input modalities	text · image · document
Output modalities	text
Provider docs	—

Gemini 3.1 Flash Lite Preview

Field	Value
Key	gemini-3.1-flash-lite-preview
Capabilities	chat · tools
Context window	1 000 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Gemini 2.5 Flash Lite

Field	Value
Key	gemini-2.5-flash-lite
Capabilities	chat · tools
Context window	1 000 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Lowest-cost Google text model. Ideal for batch workloads or simple extraction lenses.

---

Google — Generative Media

Imagen 4

Field	Value
Key	imagen-4
Capabilities	image_generation
Input modalities	text
Output modalities	image
Provider docs	—

Synchronous. High-fidelity photorealistic image generation.

Veo 3

Field	Value
Key	veo-3
Capabilities	video_generation
Input modalities	text
Output modalities	video
Provider docs	—

Async video generation (task-poll pattern). Produces cinematic-quality clips.

Lyria 2

Field	Value
Key	lyria-2
Capabilities	audio_generation · music_generation
Input modalities	text
Output modalities	audio
Provider docs	—

Async music synthesis. Outputs full instrumental tracks from a text prompt.

---

Mistral

Mistral Large 3

Field	Value
Key	mistral-large-3
Capabilities	chat · tools · json_schema
Context window	128 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Mistral's flagship instruction model. Strong structured-output and tool-use performance.

Magistral Medium 1.2

Field	Value
Key	magistral-medium-1.2
Capabilities	chat · reasoning
Context window	40 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

Magistral Small 1.2

Field	Value
Key	magistral-small-1.2
Capabilities	chat · reasoning
Context window	40 000 tokens
Input modalities	text
Output modalities	text
Provider docs	—

---

Stability AI

Stable Diffusion 4

Field	Value
Key	stable-diffusion-4
Capabilities	image_generation
Input modalities	text · image
Output modalities	image
Provider docs	—

Accepts an optional reference image for image-to-image workflows. Synchronous generation.

---

ElevenLabs

ElevenLabs v4

Field	Value
Key	elevenlabs-v4
Capabilities	audio_generation
Input modalities	text
Output modalities	audio
Provider docs	—

High-quality text-to-speech with voice cloning. Returns an audio file via the task-poll pattern.

---

Kling

Kling 2.0

Field	Value
Key	kling-2.0
Capabilities	video_generation
Input modalities	text
Output modalities	video
Provider docs	—

Async video generation. Strong at character-consistent motion.

---

Suno

Suno v5

Field	Value
Key	suno-v5
Capabilities	audio_generation · music_generation
Input modalities	text
Output modalities	audio
Provider docs	—

Produces full songs (vocals + instrumentation) from a prompt. Async; uses the task-poll pattern.

---

Midjourney

Midjourney 7

Field	Value
Key	midjourney-7
Capabilities	image_generation
Input modalities	text
Output modalities	image
Provider docs	—

Premium artistic image generation. Not yet active in the default registry; enable via BYOK.

---

Using a model in a lens

Reference a model by its key in the lens version's model_id field or pass it in the execution DTO:

{
  "model_id": "gemini-2.5-pro",
  "input_snapshot": { "prompt": "Explain quantum entanglement simply." },
  "funding_source": "platform_credit"
}

Generative media lenses declare their output in output_contract.kind (image, video, audio, music). The execution engine routes to the correct GenerativeMediaAdapter automatically.