OrcaRouter Releases OrcaCyber Zero 1.5 With 1M Context for Cybersecurity Research

A new cybersecurity model from OrcaRouter

OrcaRouter has released OrcaCyber Zero 1.5, a gated AI model designed for authorized security research. The model posts on its successor, OrcaCyber Zero 1.0, which shipped on September 17, 2026 ¹. It is post-trained for vulnerability reproduction, exploit development, penetration testing, and cyber reasoning across large codebases.

The model carries a 1M-token context window — meaning it can read very long documents or codebases in one go — and supports up to 128K tokens of output ². It features native function calling, which lets the model call tools and APIs directly, and structured outputs for clean data responses.

Benchmark scores and limits

All published scores come directly from the vendor and have not been independently verified ³. The model achieved 100% on Cybench, a set of 39 professional capture-the-flag cybersecurity tasks ¹. On CVE-Bench, built from critical web vulnerabilities, it scored 95.8% across an evaluable subset of 24 out of 40 tasks ². It also scored 93.9% on HumanEval+, a coding challenge set, and 76.5% on SWE-bench Pro V2, a software engineering benchmark .

The parameter count was not disclosed, and the model is only available through OrcaRouter's hosted API with no open weights ¹.

Priced far below rival frontier models

Pricing sits at $3.00 per 1M input tokens and $7.50 per 1M output tokens, with prompt caching costing $0.75 per 1M tokens ². For comparison, Anthropic's Claude Mythos Preview lists at $25 per 1M input and $125 per 1M output after credits ¹.

Access requires approval through a Security Research tier. Users must provide an engagement description, a passkey, and accept terms of service ². The tier targets trusted red teams and authorized security researchers.

Designed for autonomous security agents

Orca says the model is built for autonomous agent workflows. It can reason through attack paths, challenge its own hypotheses, and rank flaws by how likely they are to be exploitable ³. Targets include remote code execution (RCE), sandbox escapes, authentication bypasses, privilege escalation, and chained attack sequences ¹.

Recent latency measurements showed a median time-to-first-token of 3.16 seconds and a 95th-percentile of 8.88 seconds over seven days, across nearly 2.4 billion tokens of traffic ². The company also announced an update to OrcaVerify Text 1.0 alongside the launch .

Key facts

• Released October 10, 2026 by OrcaRouter
• 1M-token context window, 128K max output
• 100% on Cybench, 95.8% on CVE-Bench subset, 76.5% on SWE-bench Pro V2
• Priced at $3/$7.50 per 1M tokens input/output
• Gated access via Security Research tier; no open weights

Sources

¹ Marktechpost AI — OrcaRouter press coverage ² OrcaRouter official API page ³ Marktechpost AI X post

Key facts

• Released October 10, 2026 by OrcaRouter
• 1M-token context window with 128K max output
• 100% on Cybench and 95.8% on CVE-Bench subset, both vendor-reported
• Priced at $3/$7.50 per 1M input/output tokens — far below Claude Mythos at $25/$125
• Gated access for vetted researchers; no open weights available

Seen on social

We're excited to introduce OrcaCyber Zero 1.5 — our latest frontier cybersecurity model, built to discover critical vulnerabilities, investigate complex attack paths, and validate real-world exploitability. Updated weights. New benchmark results: • 95.8% — CVE-Bench (23/24…

— OrcaRouter 🐳 on X, see the post

Sources

1. OrcaRouter Releases OrcaCyber Zero 1.5 Cybersecurity Model With 1M Context — marktechpost.com
2. OrcaCyber Zero 1.5 API — Pricing & Benchmarks · OrcaRouter
3. Marktechpost AI on X: "OrcaRouter just released OrcaCyber Zero 1.5, a ...
4. OrcaRouter (@OrcaRouter) / X