Models
Six model families, one API
Use Halo 2 for most work. Reach for a specialist when the job is code, images, video or voice.
New
halo-2
Halo 2
Our flagship for long work: reasoning you can adjust, careful tool use, and a calmer default. The right first choice for almost everything.
Read the launch note
Context
400K tokens
Input
$2.00 / 1M tokens
Output
$8.00 / 1M tokens
Reasoning
Adjustable effort
Modalities
Text and image in, text out
Specialist models
halo-2-mini
Halo 2 Mini
Fast and inexpensive for classification, extraction and short replies.
Context
128K
Input
$0.25 / 1M
Output
$1.00 / 1M
forge-code-1
Forge Code 1
Tuned for long coding sessions with tools, tests and project memory.
Context
256K
Input
$1.50 / 1M
Output
$6.00 / 1M
speak-3
Speak 3
Real-time voice in and out, with natural turn-taking in forty languages.
Speech to speech
$0.06 / min
Text to speech
$12.00 / 1M chars
Transcription
from $0.12 / hr
studio-image-2
Studio Image 2
Generate and edit images from text or a reference, at 1K or 2K.
Input
Text, image
1K
$0.04 / image
2K
$0.08 / image
studio-video-1
Studio Video 1
Turn a prompt or a still into a short clip with described motion.
Input
Text, image
720p
$0.06 / sec
1080p
$0.10 / sec
Not sure where to start?
Start with Halo 2. Switch one parameter later if a specialist fits better.
Read the API overview
Which model should I choose?
Compare at a glance
Your choice depends on the job. For everything that is not audio, image or video, start with Halo 2.
halo-2
halo-2-mini
forge-code-1
studio-image-2
studio-video-1
speak-3
Model aliases
Aliases let you move to the next version of a model without a code change, or pin one release when you need consistency.
Read the changelog
halo-2
Points to the latest stable release of the model. Recommended for most apps.
halo-2-latest
Points to the newest release, including previews. Good for trying new features early.
halo-2-2026-09-21
A fixed release that never changes. Use it when results must stay identical.