- Text Generation
LFM2.5-1.2B-Thinking (free) is LiquidAI’s 1.2B-parameter, open-weight reasoning model optimized to run entirely on-device under roughly 1 GB of memory. It focuses on chain-of-thought style “thinking” before…
Powered by Google
Lyria 3 Pro Preview
- Text Generation
Lyria 3 Pro Preview is Google DeepMind’s flagship music generation model, optimized for producing full-length, structurally coherent songs from text or image prompts. It outputs high-quality 48 kHz stereo audio and is available in preview via the Gemini API and Google AI Studio.
Providers Comparison
Speed
Output tokens per second · Higher is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
End-to-End Response Time
Seconds · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
Price
USD per 1M tokens (blended) · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
About the model
What is Lyria 3 Pro Preview?
Lyria 3 Pro Preview is Google’s advanced AI music generation model designed to create full songs with complex structural coherence, including multiple verses, choruses, and bridges. It is mainly used to generate three-minute, high-fidelity stereo music tracks from text prompts or image inputs, giving creators fine-grained control over song sections and styles. It also serves developers and enterprises that need scalable music generation through platforms like Vertex AI, Gemini API, and Google AI Studio. The model follows earlier Lyria versions such as Lyria 3, extending their 30-second clip capabilities into full-length compositions within the same Lyria family.
Model capabilities
5 Core Capabilities
-
Conversational Chat
Generates context-aware, multi-turn conversational responses, handling complex instructions, following user intent, and maintaining coherent dialogue in natural language.
-
Music Generation
Produces high-quality music audio from prompts, supporting varied styles, structures, and instrumentation, optimized for creative composition and experimentation.
-
Lyric Handling
Understands and generates song lyric text aligned to musical concepts, styles, and themes for use alongside audio-based music generation workflows.
-
Multilingual Text
Processes and generates text across multiple languages, enabling cross-lingual interaction with prompts and outputs for global creative use cases.
-
Text Prompting
Interprets detailed text prompts describing musical mood, genre, and structure to guide resulting music generation outputs effectively and consistently.
Use cases
6 Most Valuable Use Cases
- Music Style Transfer
- Melody To Orchestration
- Generative Film Scoring
- Adaptive Game Soundtracks
- Advertising Jingle Creation
- AI Music Prototyping
Transparent pricing
Cost Comparison
LLM API offers the lowest cost and latency for Lyria 3 Pro–class music generation.
| Provider | Region | Latency | Throughput | Uptime | Input ($/1M) | Output ($/1M) | Context |
|---|---|---|---|---|---|---|---|
| LLM API BEST | Global | 200ms | 120 tracks/min | 99.99% | $0.40/1K tokens | $0.40/1K tokens | 32K tokens |
| Global | ~350ms | ~60 tracks/min | 99.9% | ~$0.80/1K tokens | ~$0.80/1K tokens | 32K tokens | |
| AWS Bedrock (3rd-party music model) | US East | ~420ms | ~45 tracks/min | 99.9% | ~$1.00/1K tokens | ~$1.00/1K tokens | ~16K tokens |
| Azure AI Studio (music model) | EU West | ~380ms | ~50 tracks/min | 99.9% | ~$0.90/1K tokens | ~$0.90/1K tokens | ~32K tokens |
Performance benchmarks
Technical Specifications
| Metric | Lyria 3 Pro Preview | GPT-4o (mini) | Claude 3 Haiku |
|---|---|---|---|
| Avg Latency | ~180ms | ~220ms | ~250ms |
| Context Window | 128K | 128K | 200K |
| Input Price ($/1M) | $0.20 | $0.15 | $0.25 |
| Output Price ($/1M) | $0.60 | $0.60 | $0.75 |
| Max Output Tokens | 8K | 4K | 4K |
| Throughput | ~60 tps | ~50 tps | ~45 tps |
| Uptime | 99.9% | 99.9% | 99.9% |
30-day usage via LLM API
- 3.8B
- Prompt tokens processed (30 days)
- 260M
- Completion tokens generated (30 days)
- 7.5M
- API requests served (30 days)
- 99.8%
- Avg uptime over last 30 days
Architecture & Integration
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Unified AI Routing
Automatically route each request to the best model across providers based on latency, cost, and quality—without changing your integration or redeploying.
One endpoint, any model -
Cost-Aware Controls
Set budgets, price caps, and model-level policies so LLM.API always picks the most cost-efficient option while meeting your performance and quality targets.
Optimize spend by design -
Resilient Fallbacks
Define automatic cross-provider fallbacks so outages, rate limits, or slow regions don’t break your app—responses keep flowing with no client-side logic.
Never go dark -
Deep Observability
Get unified logs, traces, and metrics across all models and vendors, with per-request insights for latency, errors, and spend in one place.
See every token -
Task-Level Orchestration
Express work as high-level tasks—like RAG, agents, or tools—and let LLM.API handle prompt shaping, model selection, and retries for consistent outcomes.
Ship features, not glue -
High-Throughput Batch
Submit massive batches across providers with automatic chunking, parallelization, and retries to maximize throughput while keeping costs and rate limits under control.
Scale to millions
Decision guide
When to Use — When NOT to Use
Use it if...
- You need high-quality music generation from text prompts, focusing on melody and structure.
- You need to quickly iterate musical ideas or demos without professional composing skills.
- Your use case involves generating varied musical styles to prototype game or app soundtracks.
- You need AI-assisted music content for social media, marketing clips, or short videos.
- Your use case involves exploring AI music capabilities within Google's Gemini and generative tools ecosystem.
Avoid if...
- You need a general-purpose language model for coding, reasoning, or text-heavy workloads.
- Your workload requires enterprise-grade guarantees around content licensing, rights management, and ownership.
- You need precise control over multi-instrument arrangements comparable to professional DAW workflows.
- Your workload requires predictable latency and throughput for large-scale production audio pipelines.
- You need fully transparent, fine-grained safety controls and governance for regulated industry deployments.
FAQ
Frequently Asked Questions
-
What is Lyria 3 Pro Preview?
Lyria 3 Pro Preview is a Google large language model accessible via LLM.API for high-quality text generation and reasoning workloads.
-
What is the context window of Lyria 3 Pro Preview?
Lyria 3 Pro Preview supports up to a 128K token context window for combined prompt and response.
-
Which modalities does Lyria 3 Pro Preview support?
Lyria 3 Pro Preview currently supports text input and text output only through LLM.API.
-
How is Lyria 3 Pro Preview priced on LLM.API?
Lyria 3 Pro Preview uses a pay-per-token pricing model; check your LLM.API dashboard for current input and output token rates.
-
How fast is Lyria 3 Pro Preview for typical requests?
For short prompts, Lyria 3 Pro Preview usually responds within a few seconds, depending on prompt size and LLM.API load.
-
What is Lyria 3 Pro Preview best suited for?
Lyria 3 Pro Preview is best for complex reasoning, code assistance, long-form content generation, and multi-step data processing tasks.
-
How do I call Lyria 3 Pro Preview via LLM.API?
Specify the provider as "google" and the model name "lyria-3-pro-preview" in your LLM.API completion or chat request.
-
How does Lyria 3 Pro Preview compare to similar Google models?
Compared to lighter Google models, Lyria 3 Pro Preview generally offers stronger reasoning and coding quality at higher compute cost.
-
Does Lyria 3 Pro Preview support streaming responses on LLM.API?
Yes, you can enable streaming in your LLM.API request to receive Lyria 3 Pro Preview tokens incrementally.
-
What are the main limitations of Lyria 3 Pro Preview?
Lyria 3 Pro Preview can hallucinate facts, reflect training data biases, and should not be relied on for unsupervised high-risk decisions.
Compare
Competitive Models
- Text Generation
Anthropic Claude Haiku (Latest) is a lightweight, fast Claude family model optimized for low-latency, cost‑efficient tasks while maintaining strong language understanding. It is notable for offering…
- Text Generation
Kimi K2.5 is MoonshotAI’s flagship open-source multimodal Mixture-of-Experts model with native vision and strong agentic capabilities, designed for long-context reasoning and complex tool use.