01
What will this model cost in production?
Compare input, output, cached, and batch pricing before a team commits to a model family.
AI model reference database
Look up AI model specs, pricing, and parameters without digging through docs.
ModelMeta turns scattered provider documentation into structured model records. People come here to check pricing, context windows, max output, supported parameters, capability flags, rate limits, and provider coverage in one place.
PricingContext windowOutput limitsParametersCapabilitiesProvider coverage
452
Model profiles with structured pricing, limits, parameters, and capability metadata.
22
Provider sources attached to the record, without turning the homepage into a provider directory.
1h
Registry refresh cadence for normalized model snapshots and route metadata.
Example normalized model record
The first screen should show the structure of a model record, not just feature one model.
A good model record combines commercial terms, token limits, runtime controls, capability flags, and source coverage in one clean page.
Input price
$2.50
sample normalized field
Output price
$7.50
sample normalized field
Context window
2M
sample token limit
Max output
32K
sample token limit
Supported parameters
temperaturetop_ptop_kreasoning_effort
Capability flags
VisionTool useStreamingStructured output
Source layer
OpenAISiliconFlowInfini-AICohereAlibaba CloudByteDance
Provider pages remain useful as provenance and coverage context, but the model record stays at the center of the product.
What this product helps answer
Teams usually need a few concrete answers before they pick a model.
People usually arrive with a practical question. The homepage should expose those questions first, then show the field groups that answer them.
02
Can it fit the workload and token limits?
Context windows, max output, modality support, and rate limits should be visible before implementation starts.
03
Which controls and capabilities are actually supported?
Teams need explicit fields for parameters, structured output, tool use, streaming, and reasoning behavior.
04
Which providers expose the model and where did the data come from?
Provider coverage stays attached to the record as source metadata instead of replacing the model as the main object.
Structured field groups
The database should make the important model fields visible in seconds.
Pricing
Pricing
Commercial terms belong near the top so users can filter quickly and compare like-for-like.
Input price
Output price
Cached input
Batch pricing
Context and limits
Context and limits
Token limits decide what fits into a workflow before anyone builds around a model.
Context window
Max input
Max output
Rate limits
Parameters
Parameters
Runtime controls should be listed explicitly, not buried inside prose or provider-specific docs.
temperature
top_p
top_k
reasoning_effort
Capabilities
Capabilities
Structured flags make model behavior easier to compare across providers and model families.
Vision
Tool use
Streaming
Structured output
Product flow
Built like a reference workflow, not a single overloaded page.
01
Browse the catalog
Search and filter by the model fields people actually compare, not by a homepage state machine.
02
Open structured profiles
Each model gets its own page for pricing, limits, capabilities, parameters, and reference metadata.
03
Compare a shortlist
Shortlists should turn into side-by-side comparisons instead of more tabs across provider docs.
Provider coverage
Providers stay visible as source context, not as the homepage headline.
OpenAI
Provider profile and coverage index
139
models
SiliconFlow
High-Performance AI Inference Platform
100
models
Infini-AI
Infinigence AI - AI Computing Optimization Platform
57
models
Cohere
Enterprise-Ready RAG & Tool Use Specialist
27
models
Provider pages explain provenance, access points, and coverage. The homepage stays centered on the model information people compare when making a decision.
Open provider directory >Featured model profiles
A few real model pages make the database tangible.
Google
FlagshipNew
Gemini 2.5 Pro
Enhanced mid-generation model with improved capabilities
Input
$2.5
per 1M tokens
Output
$7.5
per 1M tokens
Context
2M
window
Max output
Not listed
tokens
Open model profile >
Google
FlagshipNew
Gemini 3 Pro
Most capable Gemini model with enhanced reasoning and multimodal understanding
Input
$3.5
per 1M tokens
Output
$10.5
per 1M tokens
Context
2M
window
Max output
8.2K
tokens
Supported controls
temperaturetop_pfrequency_penalty
Open model profile >
Google
FlagshipNew
Gemini 2.0 Flash
Next generation features, superior speed, native tool use, and multimodal generation
Input
$0.075
per 1M tokens
Output
$0.3
per 1M tokens
Context
1M
window
Max output
Not listed
tokens
Open model profile >