Stay Frontier.
Afoim gives developers and enterprises a single, OpenAI-compatible API for the world's leading large models: Anthropic's Claude, OpenAI's GPT, Google's Gemini, DeepSeek, Kimi, GLM, and Afoim's own researched models. One API key, one bill, smart routing and automatic failover across providers, plus our in-house model lab pushing the frontier itself.
No credit card required for the sandbox · SOC 2 aligned controls · 99.9% uptime SLA on Enterprise plans
from openai import OpenAI client = OpenAI( base_url="https://api.afoim.com/v1", api_key="sk-afoim-…", # ONE key ) # Switch models with one parameter: # "claude-…" · "gpt-…" · "gemini-…" # "deepseek-…" · "kimi-…" · "glm-…" # "orrery-v3.1-flash" (our own flagship) resp = client.chat.completions.create( model="orrery-v3.1-flash", messages=[{"role": "user", "content": "Summarize this contract…"}], )
OpenAI-compatible endpoint · Streaming · Tool use · Smart routing & failover · Unified billing
Selected customers and research partners. Read the results →
The leading models, behind one key
Stop juggling six provider accounts, six SDKs, and six invoices. Afoim's gateway routes one OpenAI-compatible API to the leading model providers, and to our own models, with enterprise controls built in.
One key, every model
A single Afoim key unlocks Claude, GPT, Gemini, DeepSeek, Kimi, GLM, and Afoim's own models. Switch models by changing one parameter, with no new contracts and no new integrations.
Drop-in compatible
Point the official OpenAI SDK at api.afoim.com/v1 and keep your existing code. Streaming, tool use, and structured outputs work across providers through one interface.
One bill, full visibility
One monthly USD invoice for all providers, with per-model, per-key, and per-team usage analytics and spend limits. No more reconciling six statements.
Routing that saves money
Route each request to the best model for the job by cost, latency, or quality, with rules you control, or let Afoim's router decide per request.
Failover, automatically
If a provider has an outage or rate-limit spike, traffic shifts to an equivalent model in milliseconds. Your users never see a provider's bad day.
Control & compliance
SSO/SAML, per-key permissions and budgets, data-residency and zero-retention options, audit logs, and dedicated support with a 99.9% SLA.
The frontier, on one platform
Provider models available through the unified Afoim API today, accessed under one key and one bill:
Claude family
Anthropic's Claude models for reasoning, coding, and long-context analysis.
GPT family
OpenAI's GPT models for general-purpose chat, agents, and multimodal work.
Gemini family
Google's Gemini models for multimodal understanding and long-context tasks.
DeepSeek family
DeepSeek's high-efficiency reasoning and coding models.
Kimi family
Moonshot AI's Kimi models, known for very long context and Chinese-language strength.
GLM family
Zhipu AI's GLM models for bilingual reasoning, coding, and agent tasks.
Claude, GPT, Gemini, DeepSeek, Kimi, and GLM are trademarks of their respective owners (Anthropic, OpenAI, Google, DeepSeek, Moonshot AI, and Zhipu AI). Afoim is an independent API platform and is not affiliated with, endorsed by, or sponsored by any of them; provider models are accessed through authorized channels and remain subject to each provider's terms.
And our own: the Afoim model family
Business two: models researched, trained, and served by Afoim itself, available through the same key, often at a fraction of frontier prices.
Orrery-v3.1-flash
Our flagship high-speed model, designed, trained, and served entirely by Afoim. Orrery-v3.1-flash is built for latency-sensitive production work: real-time copilots, support automation, and high-volume generation where response time decides the experience. It is called through the same unified API, using the model name orrery-v3.1-flash, under the same key and the same bill as every other model on the platform. Specifications: 1M context window, image input supported, streaming and tool use supported. Pricing: $0.25 per 1M input tokens and $0.45 per 1M output tokens. The full Orrery family has three models: Orrery-v3.1-flash and Orrery-v3.0-flash for fast, image-aware work, and Orrery-v3.0-pro, our flagship reasoning and long-form model. All three carry a 1M context window.
| Afoim model | Best for | Context | Input / 1M tokens | Output / 1M tokens |
|---|---|---|---|---|
| Orrery-v3.1-flash | Flagship speed: real-time copilots, support automation, high-volume generation. Image input supported | 1M | $0.25 | $0.45 |
| Orrery-v3.0-flash | Fast, economical generation and image understanding at scale. Image input supported | 1M | $0.20 | $0.40 |
| Orrery-v3.0-pro | Flagship reasoning and long-form work: complex agents, advanced coding, book-length documents | 1M | $2.50 | $12.00 |
Provider-model pricing passes through at published rates plus a transparent platform margin shown before you call, with no surprises. Prompt caching and smart routing typically cut effective spend 20–40%.
Research at the frontier, published in the open
Our Oregon-based research team works on the problems that decide whether large models are trustworthy: alignment, interpretability, evaluation, and efficient training. We publish what we learn.
Alignment & safety
Constitutional training methods, red-teaming, and refusal calibration, so capability gains never outrun control.
Interpretability
Opening the black box: feature-level analysis of what our models represent and why they answer the way they do.
Efficiency
Distillation, sparsity, and inference optimization that cut cost per token, with savings passed straight to API pricing.
What builders achieve with Afoim
Northwind Labs: SaaS support copilot
Northwind routes 40,000 tickets a month through Orrery-v3.1-flash with tool use into its help desk. Resolution time fell from 9 hours to 22 minutes while CSAT rose 11 points.
Meridian Legal: contract analysis
Meridian's attorneys use Orrery-v3.0-pro to first-pass review commercial contracts, flagging risk clauses with cited evidence. Partner review time dropped from days to hours.
Cascade Financial: document intelligence
Cascade processes loan documents with structured outputs on Afoim, replacing a legacy OCR pipeline and cutting manual correction by 87%.
Helix Bio: literature synthesis
Helix Bio's scientists query 2M+ papers through Afoim-powered retrieval, compressing literature reviews from weeks into days.
Built like infrastructure. Run like a partner.
Our slogan, and the standard behind it
Our company line, introduced at our founding in November 2026, is:
- “Stay Frontier.” (our primary slogan)
- “Make Intelligence Limitless.”
- “Make Every Model Yours.”
- “Make Progress Without Limits.”
The first line is the one we print on the building. The others are alternates we keep for campaigns. Tell us your favorite via the contact form.
Reliability first
Multi-region serving, automatic failover, transparent status, and an SLA with teeth. If we miss 99.9%, you get credits automatically.
Your data stays yours
API inputs and outputs are never used to train our models by default. Zero-retention arrangements are available for regulated industries.
Humans who answer
Real engineers on support, a named account team on Enterprise plans, and a phone number, +1 (971) 521-0250, that reaches a person.
Start building today.
$25 in starter credits, a sandbox in under five minutes, and docs written by the people who built the models.
Get API access