ares-1 · now generally available

Frontier intelligence,
made abundant.

ares-1 matches the world's best models on every benchmark that matters, and runs at the speed and cost of yesterday's open-source. Build the systems you've been pricing out of existence. Sign up, drop in an API key, ship in 60 seconds.

0%
SWE-bench Verified
$0
Per 1M input tokens
0
Median throughput
0ms
Time to first token
Powering builders at
Lattice Orbital Verge Systems Northwind Halo Labs Cobalt
01 — Benchmarks

Beats the giants on the tests that matter.

ares-1 was trained for real engineering work — long-context reasoning, tool use, and code. Independently evaluated against the latest frontier models.

SWE-bench Verified
Resolved real GitHub issues, single-shot pass@1
02 — Pricing

Built so you can use it everywhere.

A price floor low enough to run intelligence on every event, every user, every line of code — not just the ones you can afford to. No commits, no tiers, no negotiation.

ares-1 — input
$0.40/ 1M tok
output · $1.60 / 1M tok
Industry average — input
$2.10/ 1M tok
output · $8.40 / 1M tok
Estimate your monthly bill
50M tokens / month
ares-1
$50.00
opus 4.7 (est.)
$262.50
gpt 5.5 (est.)
$235.00
03 — Deploy

Live in sixty seconds.

Drop-in compatible with the OpenAI SDK. No new client libraries, no vendor lock-in. Sign up with a Google account and start streaming.

STEP 01

Sign up

Auth with Google, GitHub, or email. $5 in free credits, no card required.

ares.dev/signup
STEP 02

Generate a key

Scoped, rotatable, and rate-limited. Org-level and per-environment.

$ export ARES_API_KEY=sk-…
STEP 03

Make your first call

OpenAI-compatible. Point your existing SDK at our base URL.

$ curl ares.dev/v1/chat...

Built for coding agents.

ares-1 ships with native tool calling, 1M-token context, parallel function execution, and structured outputs. Used in production today by the leading IDE assistants and autonomous engineering platforms.

  • OpenAI-compatible /v1/chat/completions
  • Native function calling, streaming, JSON mode
  • Python, Node, Go, Rust, and Ruby SDKs
  • 1M token context, prompt caching, batch API
04 — Enterprise

Built for the strictest environments.

Every request, by default, is private. No training on customer data. No retention beyond the lifetime of the request. SOC 2 Type II, HIPAA, and EU-resident deployments.

Zero Data Retention

Inputs and outputs are discarded the moment a request completes. Not logged, not cached, not retained. Verifiable via audit log.

DEFAULT · ALL TIERS

No Training on Your Data

Customer prompts and completions are contractually excluded from training corpora. Indemnified for IP and confidentiality.

CONTRACTUAL GUARANTEE

SOC 2 Type II · HIPAA

Audited annually by Coalfire. BAA available. ISO 27001 in progress. Penetration tested quarterly by third-party.

REPORT ON REQUEST

EU & US Data Residency

Pin inference to Frankfurt, Dublin, or Ashburn. GDPR-compliant by construction. DPA + SCCs available.

REGIONAL PINNING

99.95% Uptime SLA

Multi-region failover. Capacity-reserved tiers. Real-time status page with incident history and root-cause reports.

CREDIT-BACKED

SSO · SCIM · Audit Logs

Okta, Microsoft Entra, Google Workspace. Per-key rate limits. Streaming audit logs to your SIEM. RBAC for orgs & projects.

ENTERPRISE TIER

Build what wasn't possible yesterday.

The teams shipping the next generation of software are building on ares. Join them — $5 in credits, no card, no waitlist.