At a Glance
| Vendor | Anthropic |
| Family | Claude Mythos (preview) |
| Launched | Anthropic previewed Claude Mythos (internal codename Capybara) in April 2026, describing it as a step change above Claude Opus 4.6 with notably stronger reasoning, coding, and cybersecurity vulnerability detection. Access is gated through Project Glasswing, a partner program limited to roughly 50 organizations. |
| Context window | Anthropic has not published full spec; preview-tier access reports indicate context handling consistent with the 4.x Opus family (200K standard with extended-context variants). |
| Pricing | Preview-tier access does not have published pricing. Anthropic's pattern (Opus 4.6 to Opus 4.7) suggests Mythos will price at or above current Opus rates when it reaches general availability, with extended-thinking workloads carrying premium inference cost. |
| Access channels | Project Glasswing partners only at preview. Roadmap pattern from prior Anthropic releases suggests broader API access via Anthropic, AWS Bedrock, Google Cloud Vertex AI, and Microsoft Foundry once general availability lands. |
Notable Benchmarks
Reported step-change gains on competition math, GPQA Diamond, SWE-bench Verified, and cybersecurity benchmarks (vulnerability detection, exploit reasoning). Anthropic has emphasized cybersecurity capability as a distinguishing axis, suggesting Mythos is the first frontier model positioned explicitly for security workloads.
Strengths
Step-change frontier reasoning, distinctive cybersecurity capability, gated access creating early-adopter signal among partner organizations, alignment with Anthropic's safety-research positioning.
Limitations
Inaccessible to most teams during preview. Not yet in general production deployments. Public benchmark numbers are pre-release and not independently replicated. Anthropic pricing pattern means Mythos will be expensive at GA.
Brand-Visibility Implications
Mythos matters for brand visibility for two reasons even before general availability. First, the partner organizations in Project Glasswing are exactly the high-signal enterprise buyers (defense, fintech, security software, frontier-AI labs) that influence broader market adoption. Second, the cybersecurity emphasis means Mythos is the model that will run unattended security audits and vendor-risk assessments. Brands flagged by Mythos as low-trust (weak signing, missing security disclosures, poor SBOM coverage) will see ripple effects in enterprise procurement workflows. See Claude visibility and Claude Opus 4.7 brief.
How Presenc AI Tracks This Model
Presenc AI monitors brand visibility on Anthropic's Claude Mythos (preview) as part of continuous multi-platform AI visibility tracking. We sample Anthropic Claude Mythos across representative prompt sets daily, compare against competitor performance on the same prompts, and flag material mention-rate changes so brand teams can respond quickly when AI representation shifts.