Download the PHP package luminovang/php-ai-models without Composer

On this page you can find all versions of the php package luminovang/php-ai-models. It is possible to download/install these versions without Composer. Possible dependencies are resolved automatically.

FAQ

After the download, you have to make one include require_once('vendor/autoload.php');. After that you have to import the classes with use statements.

Example:
If you use only one package a project is not needed. But if you use more then one package, without a project it is not possible to import the classes with use statements.

In general, it is recommended to use always a project to download your libraries. In an application normally there is more than one library needed.
Some PHP packages are not free to download and because of that hosted in private repositories. In this case some credentials are needed to access such packages. Please use the auth.json textarea to insert credentials, if a package is coming from a private repository. You can look here for more information.

  • Some hosting areas are not accessible by a terminal or SSH. Then it is not possible to use Composer.
  • To use Composer is sometimes complicated. Especially for beginners.
  • Composer needs much resources. Sometimes they are not available on a simple webspace.
  • If you are using private repositories you don't need to share your credentials. You can set up everything on our site and then you provide a simple download link to your team member.
  • Simplify your Composer build process. Use our own command line tool to download the vendor folder as binary. This makes your build process faster and you don't need to expose your credentials for private repositories.
Please rate this library. Is it a good library?

Example code of luminovang/php-ai-models

Informations about the package php-ai-models

PHP AI Model Helper Class

Type: const Model: string (final-class)

A static catalogue of AI model identifiers. Constants eliminate typos, enable IDE autocompletion, and create a single source across codebase.

Installation

Luminova\AI\Model (enum version) — PHP string-backed enum class.

Table of Contents

Overview

Model is a final class that cannot be instantiated. All constants and methods are accessed statically:

Import

No additional dependencies. The class uses ReflectionClass (PHP core) only for the all() helper method.

Naming Convention

Rule Example
Hyphens and dots → underscores gpt-4.1-mini → GPT_4_1_MINI
Size tag suffix (:8b) llama3.1:8b → LLAMA_3_1_8B
MoE tag (8x7b) mixtral:8x7b → MIXTRAL_8X7B
Versioned snapshot claude-opus-4-5-20251101 → CLAUDE_OPUS_4_5_SNAP
Clean alias alongside snapshot claude-opus-4-5 → CLAUDE_OPUS_4_5

Constants Reference

OpenAI — GPT-5 Family

Constant API Value Notes
Model::GPT_5 gpt-5 Flagship model. Complex reasoning, multimodal, 256 K context.
Model::GPT_5_MINI gpt-5-mini Faster, more affordable GPT-5 variant.
Model::GPT_5_NANO gpt-5-nano Smallest GPT-5; optimized for latency and cost.

OpenAI — GPT-4.1 Family

Constant API Value Notes
Model::GPT_4_1 gpt-4.1 1 M token context, instruction-following, coding. Supports fine-tuning.
Model::GPT_4_1_MINI gpt-4.1-mini Default chat model for the Luminova OpenAI client. Supports fine-tuning.
Model::GPT_4_1_NANO gpt-4.1-nano Fastest / cheapest GPT-4.1. Supports fine-tuning.

OpenAI — GPT-4o Family

Constant API Value Notes
Model::GPT_4O gpt-4o Multimodal (text + image + audio). 128 K context.
Model::GPT_4O_MINI gpt-4o-mini Lightweight GPT-4o. 128 K context.
Model::GPT_4O_AUDIO gpt-4o-audio-preview Native audio I/O.
Model::GPT_4O_MINI_AUDIO gpt-4o-mini-audio-preview Lower-cost audio variant.
Model::GPT_4O_REALTIME gpt-4o-realtime-preview Low-latency real-time speech and text.
Model::GPT_4O_MINI_REALTIME gpt-4o-mini-realtime-preview Lower-cost realtime variant.
Model::COMPUTER_USE computer-use-preview GUI interaction via the Responses API.

OpenAI — Reasoning (o-series)

Constant API Value Notes
Model::O3 o3 Most capable reasoning model. Supports visual reasoning.
Model::O3_PRO o3-pro o3 with extra compute for critical tasks.
Model::O3_DEEP_RESEARCH o3-deep-research Multi-step web and document research.
Model::O4_MINI o4-mini Fast reasoning; top benchmark for math/coding/vision.
Model::O4_MINI_DEEP_RESEARCH o4-mini-deep-research Deep research variant of o4 Mini.

OpenAI — Image Generation

Constant API Value Notes
Model::GPT_IMAGE_1_5 gpt-image-1.5 Latest image model. High-resolution + inpainting. Requires approval.
Model::GPT_IMAGE_1 gpt-image-1 Default image model for the Luminova OpenAI client. Requires approval.
Model::DALL_E_3 dall-e-3 Generally available. Up to 1792×1024 px.
Model::DALL_E_2 dall-e-2 Previous generation; lower cost.

OpenAI — Text-to-Speech

Constant API Value Notes
Model::GPT_4O_MINI_TTS gpt-4o-mini-tts Default TTS model. Voices: alloy, echo, fable, onyx, nova, shimmer.
Model::TTS_1 tts-1 Optimized for real-time use.
Model::TTS_1_HD tts-1-hd Higher quality, more natural intonation.

OpenAI — Transcription

Constant API Value Notes
Model::GPT_4O_TRANSCRIBE gpt-4o-transcribe Superior accuracy, multilingual.
Model::GPT_4O_MINI_TRANSCRIBE gpt-4o-mini-transcribe Faster, lower-cost. Currently recommended.
Model::WHISPER_1 whisper-1 Default transcription model. 99+ languages.

OpenAI — Embeddings

Constant API Value Notes
Model::TEXT_EMBEDDING_3_LARGE text-embedding-3-large Highest accuracy. 3072-dimensional (reducible). Best for RAG.
Model::TEXT_EMBEDDING_3_SMALL text-embedding-3-small Default embedding model. 1536-dimensional.
Model::TEXT_EMBEDDING_ADA_002 text-embedding-ada-002 Legacy. Prefer TEXT_EMBEDDING_3_SMALL for new work.

OpenAI — Moderation

Constant API Value Notes
Model::OMNI_MODERATION omni-moderation-latest Text + image moderation.
Model::TEXT_MODERATION text-moderation-latest Text-only moderation.

Claude (Anthropic) — 4.6 Generation (current)

Constant API Value Notes
Model::CLAUDE_OPUS_4_6 claude-opus-4-6 Most capable. ~14.5 h task horizon. 1 M context (beta).
Model::CLAUDE_SONNET_4_6 claude-sonnet-4-6 Default Claude model. Preferred by developers over previous Opus.

Claude (Anthropic) — 4.5 Generation

Constant API Value Notes
Model::CLAUDE_OPUS_4_5 claude-opus-4-5 67% price cut, 76% fewer output tokens vs previous Opus.
Model::CLAUDE_OPUS_4_5_SNAP claude-opus-4-5-20251101 Pinned snapshot — guaranteed reproducibility.
Model::CLAUDE_SONNET_4_5 claude-sonnet-4-5 Industry-leading agent capabilities.
Model::CLAUDE_HAIKU_4_5 claude-haiku-4-5 Fastest, most cost-effective Claude 4.5.
Model::CLAUDE_HAIKU_4_5_SNAP claude-haiku-4-5-20251001 Pinned snapshot.

Claude (Anthropic) — 4.1 Generation

Constant API Value Notes
Model::CLAUDE_OPUS_4_1 claude-opus-4-1 Industry leader for coding and long-horizon agentic tasks.
Model::CLAUDE_OPUS_4_1_SNAP claude-opus-4-1-20250805 Pinned snapshot.
Model::CLAUDE_SONNET_4_1 claude-sonnet-4-1 Production-ready agents at scale.

Claude (Anthropic) — 4.0 Generation

Constant API Value Notes
Model::CLAUDE_OPUS_4 claude-opus-4-0 First Claude 4-gen Opus. State-of-the-art coding at release.
Model::CLAUDE_SONNET_4 claude-sonnet-4-0 First Claude 4-gen Sonnet. Fast and context-aware.

Claude (Anthropic) — 3.7 Generation

Constant API Value Notes
Model::CLAUDE_SONNET_3_7 claude-sonnet-3-7 Introduced extended (hybrid) thinking.
Model::CLAUDE_SONNET_3_7_SNAP claude-3-7-sonnet-20250219 Pinned snapshot.

Claude (Anthropic) — 3.5 Generation (legacy)

Constant API Value Notes
Model::CLAUDE_SONNET_3_5 claude-3-5-sonnet-20241022 Upgraded Sonnet with computer use (Oct 2024).
Model::CLAUDE_HAIKU_3_5 claude-3-5-haiku-20241022 Lightweight, fast. Ideal for rapid completions.

Ollama — Llama Family (Meta)

Constant API Value Notes
Model::LLAMA_3 llama3 Baseline Llama 3 (8 B). Most widely deployed.
Model::LLAMA_3_1 llama3.1 128 K context support.
Model::LLAMA_3_1_8B llama3.1:8b Explicit 8 B tag.
Model::LLAMA_3_1_70B llama3.1:70b Large-scale; multi-GPU or high-VRAM.
Model::LLAMA_3_2 llama3.2 Compact (1 B / 3 B). Optimized for edge hardware.
Model::LLAMA_3_2_1B llama3.2:1b Ultra-compact for edge and embedded use.
Model::LLAMA_3_2_3B llama3.2:3b Small but capable for CLI copilots.
Model::LLAMA_3_3 llama3.3 Latest large Llama (70 B). Excellent long-form chat.
Model::LLAMA_3_3_70B llama3.3:70b Explicit 70 B tag.

Ollama — Gemma Family (Google)

Constant API Value Notes
Model::GEMMA_3 gemma3 Current-gen (1 B–27 B). 128 K context; vision-capable (4 B+).
Model::GEMMA_3_4B gemma3:4b Vision-capable; fits 8 GB VRAM.
Model::GEMMA_3_12B gemma3:12b 12–16 GB VRAM sweet spot.
Model::GEMMA_3_27B gemma3:27b Flagship Gemma 3 variant.
Model::GEMMA_2 gemma2 Previous gen; proven reliability (2 B, 9 B, 27 B).
Model::GEMMA_2_2B gemma2:2b Smallest Gemma 2; edge deployments.
Model::GEMMA_2_9B gemma2:9b Good performance within 10 GB VRAM.
Model::GEMMA_2_27B gemma2:27b Creative and NLP-focused tasks.

Ollama — Mistral / Mixtral

Constant API Value Notes
Model::MISTRAL mistral Fast 7 B model with strong European language support.
Model::MISTRAL_7B mistral:7b Explicit 7 B tag.
Model::MIXTRAL_8X7B mixtral:8x7b Mixture-of-Experts; 2 experts active per token.
Model::MIXTRAL_8X22B mixtral:8x22b Larger MoE; near-frontier quality for local hardware.

Ollama — Qwen Family (Alibaba)

Constant API Value Notes
Model::QWEN_3 qwen3 Latest generation. Up to 256 K context; strong multilingual.
Model::QWEN_3_4B qwen3:4b Compact; fits low-VRAM hardware.
Model::QWEN_3_14B qwen3:14b Mid-range; single consumer GPU.
Model::QWEN_3_72B qwen3:72b Maximum capability; enterprise-grade.
Model::QWEN_2_5 qwen2.5 Previous gen; 18 T tokens; 128 K context.
Model::QWEN_2_5_7B qwen2.5:7b
Model::QWEN_2_5_14B qwen2.5:14b
Model::QWEN_2_5_CODER qwen2.5-coder Coding-focused; 87 languages; matches GPT-4o at 32 B.
Model::QWEN_2_5_CODER_7B qwen2.5-coder:7b Excellent code quality on limited hardware.
Model::QWEN_2_5_CODER_32B qwen2.5-coder:32b Best local coding model at this scale.

Ollama — DeepSeek Family

Constant API Value Notes
Model::DEEPSEEK_R1 deepseek-r1 Open reasoning model; matches o3 on key benchmarks.
Model::DEEPSEEK_R1_7B deepseek-r1:7b Smallest R1; 8–10 GB VRAM.
Model::DEEPSEEK_R1_14B deepseek-r1:14b Best mid-range reasoning for home labs.
Model::DEEPSEEK_R1_32B deepseek-r1:32b 24 GB+ VRAM setups.
Model::DEEPSEEK_R1_70B deepseek-r1:70b Near-frontier; multi-GPU recommended.
Model::DEEPSEEK_CODER deepseek-coder 87 programming languages; 2 T training tokens.
Model::DEEPSEEK_CODER_33B deepseek-coder:33b Top-quality local code generation.

Ollama — Phi Family (Microsoft)

Constant API Value Notes
Model::PHI_4 phi4 Latest lightweight model; 14 B, 128 K context.
Model::PHI_4_14B phi4:14b Explicit 14 B tag.
Model::PHI_3 phi3 Previous gen (3.8 B Mini / 14 B Medium).
Model::PHI_3_MINI phi3:mini 3.8 B; suitable for on-device and IoT.

Ollama — Coding Models

Constant API Value Notes
Model::CODE_LLAMA codellama Meta's code-focused Llama (7 B–70 B). Fill-in-the-middle support.
Model::CODE_LLAMA_13B codellama:13b Good balance of code quality and hardware.
Model::CODE_LLAMA_34B codellama:34b High-quality generation for 24 GB VRAM.

Ollama — Vision Models

Constant API Value Notes
Model::LLAVA llava Default vision model for the Luminova Ollama client.
Model::LLAVA_13B llava:13b Stronger vision understanding.
Model::LLAVA_34B llava:34b Highest-quality LLaVA; 24+ GB VRAM.
Model::LLAMA_3_2_VISION llama3.2-vision Better structured-output than LLaVA.
Model::MOONDREAM moondream Tiny (1.8 B); edge devices; fast captioning.
Model::BAKLLAVA bakllava Mistral-7B base with LLaVA multimodal fine-tuning.

Ollama — Embedding Models

Constant API Value Notes
Model::NOMIC_EMBED_TEXT nomic-embed-text Default embedding model. 8 K context; strong MTEB scores.
Model::MXBAI_EMBED_LARGE mxbai-embed-large 1024-dimensional; competitive with OpenAI's large model.
Model::ALL_MINILM all-minilm 384-dimensional; very fast similarity search.

Static Methods

client(string $model): string|null

Return the client short-name for a given model string. Returns null for unknown models.

forClient(string $client): array

Return all ['CONST_NAME' => 'model-id'] pairs that belong to a specific client.

forCapability(string $capability): array

Return all ['CONST_NAME' => 'model-id'] pairs that support a given capability tag.

Available tags: chat, vision, image, embedding, speech, transcription, reasoning, coding, fine-tuning, moderation.

capabilities(string $model): array

Return all capability tags for a given model string.

isVision(string $model): bool

isReasoning(string $model): bool

isEmbedding(string $model): bool

exists(string $model): bool

Check whether a model string is catalogued. Useful for validating user-supplied input before sending it to a client API.

all(): array

Return every public constant as a ['CONST_NAME' => 'model-id'] map using reflection. Private constants (PROVIDER_MAP, CAPABILITY_MAP) are automatically excluded.

Usage Examples

Basic Usage

Validating User Input

Iterating a Client's Models

Filtering by Capability

Guarding Vision Calls

Routing by Client

See Also

All versions of php-ai-models with dependencies

PHP Build Version
Package Version
Requires php Version ^8
Composer command for our command line client (download client) This client runs in each environment. You don't need a specific PHP version etc. The first 20 API calls are free. Standard composer command

The package luminovang/php-ai-models contains the following files

Loading the files please wait ...