Anthropic launches Claude Fable 5.1 and Mythos 5.1 with a tighter cost and safety profile
One base model now serves general use and trusted access, putting cost, data residency, and safeguards in the same deployment decision.
Key takeaways
Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1. They share one base model but use different safeguard and access profiles. Fable is generally available and is estimated to cost 25% less for typical token workloads, with savings of up to about 45% for highly agentic workloads. Enterprise Frontier Safeguards will keep customer data in infrastructure controlled by the customer while providing misuse detection. Mythos is offered through trusted access programs for cybersecurity and life sciences. Anthropic also described software vulnerability discovery, protein binder design, and GPU kernel optimization examples. For enterprise teams, the launch makes model selection a joint decision about capability, cost, data residency, and risk controls.
What happened
Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1. They share one base model but use different safeguard and access profiles. Fable is generally available and is estimated to cost 25% less for typical token workloads, with savings of up to about 45% for highly agentic workloads. Enterprise Frontier Safeguards will keep customer data in infrastructure controlled by the customer while providing misuse detection. Mythos is offered through trusted access programs for cybersecurity and life sciences. Anthropic also described software vulnerability discovery, protein binder design, and GPU kernel optimization examples. For enterprise teams, the launch makes model selection a joint decision about capability, cost, data residency, and risk controls.
Why it matters
Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1. They share one base model but use different safeguard and access profiles. Fable is generally available and is estimated to cost 25% less for typical token workloads, with savings of up to about 45% for highly agentic workloads. Enterprise Frontier Safeguards will keep customer data in infrastructure controlled by the customer while providing misuse detection. Mythos is offered through trusted access programs for cybersecurity and life sciences. Anthropic also described software vulnerability discovery, protein binder design, and GPU kernel optimization examples. For enterprise teams, the launch makes model selection a joint decision about capability, cost, data residency, and risk controls.
Actions for teams
- Record the source, publication date, and scope before separating facts from interpretation.
- Measure cost, permissions, logs, and escalation rates on one low-risk task.
- This AI-assisted article is checked by an automated audit for source evidence, bilingual fields, and page safety before publication.
Software catalog Learning catalog AI news catalog
Conclusion
ENHE should treat this release as a measurable operating change. Teams can adopt the same evidence-first routine for future model updates.
What this means for everyday users
ENHE 读者可把这次发布当作企业模型采购的评估模板：分别核算 token 成本、数据驻留、访问资格和护栏误报，而不是只比较榜单分数。
Tools you may use
Related tutorials
Related Tools And Tutorials
Use the following ENHE AI sections to continue from the news signal into tool selection, account-service guidance, or practical learning.
Related reading
OpenAI Daybreak Models Arrive on Amazon Bedrock with Blue and Red Access Tiers
OpenAI and AWS announced on August 11, 2026 that Daybreak Blue and Daybreak Red are available through Amazon Bedrock for defensive cybersecurity work. Organizations must register and receive approval; access is not automatically enabled for every AWS account. After approval, teams can select a model in the Bedrock console or integrate through Bedrock Mantle, which offers an interface compatible with the OpenAI Responses API. Availability does not make security analysis automatic or safe. Start with sanitized logs, a sample repository, or an isolated environment. Restrict network, credential, and shell permissions, preserve audit logs, require human approval for changes, and measure reproducible findings, false positives, missed issues, tested fixes, elapsed time, and cost before expanding scope.
How to Start an AI-Assisted Security Review: A Six-Step Read-Only Guide
OpenAI cofounder Greg Brockman published The Defender's Window on August 17, 2026, arguing that advanced AI capability should be directed toward cyber defense. For an ordinary team, the responsible starting point is not an agent that changes production. Select one repository or a sanitized log set, define a read-only permission and data boundary, inventory the assets, and write explicit threat assumptions. Require every candidate finding to include evidence and reproduction steps, then have a human classify it. Implement a proposed fix only in an isolated branch and require tests, code review, and a rollback exercise. This six-step template treats the OpenAI article as a direction, not proof that a model finding or an organization's security posture has been verified.
GitHub Copilot Adds Team-Level Model Policy Targeting for Enterprises
GitHub announced a public preview of enterprise team model policy targeting on July 31, 2026, with gradual availability for most customers around August 3. Enterprise administrators can retain an Enabled, Disabled, or Optional baseline and add optional models for selected teams. GitHub says a person who belongs to multiple teams in the same enterprise receives the least restrictive effective model access. This enables task- and budget-based rollout, but overlapping membership can create unintended permission. Export the current baseline, build a team-to-model matrix, include one overlapping test user, verify access in the model picker, IDE, CLI, and agent surfaces, monitor premium requests and model charges, inspect audit logs, and test revocation before expanding.
GitHub Enterprise Server 3.22 RC Brings Copilot CLI to Isolated Environments
GitHub released the Enterprise Server 3.22 candidate on August 11, 2026 with a technical preview that lets administrators configure Copilot CLI for enterprises operating without connectivity to GitHub Cloud. Users can authenticate with GHES credentials and use the configured model provider. GitHub's Copilot CLI documentation adds an important boundary: offline mode provides full network isolation only when the provider is also local or inside the isolated environment. If the provider base URL is remote, prompts and code context still travel to that provider. Teams should validate a test instance, model tool calling and streaming, certificates, DNS, proxies, outbound rules, repository permissions, audit logs, failure handling, upgrades, and rollback before considering production use.
Google Launches Gemini 3.7 Flash with Introductory Pricing for Coding and Agents
Google introduced Gemini 3.7 Flash on August 13, 2026 for coding, web development, complex knowledge work, and agentic tasks. The rollout covers the Gemini API, Google AI Studio, Antigravity, Android Studio, enterprise products, and Gemini Spark in eligible plans and regions; GitHub also began adding it to Copilot. Google lists introductory pricing of $0.75 per million input tokens and $3.75 per million output tokens through December 31, before planned standard pricing doubles on January 1, 2027. Treat the launch as a controlled evaluation: compare one fixed task with your current model and record quality, latency, tokens, tool calls, failures, and human rework before migrating production workflows.
GitHub Copilot Usage Metrics Adds Agent-App Activity
GitHub announced on August 7, 2026 that the Copilot Usage Metrics API now reports activity from third-party agent apps. Enterprise, organization, enterprise-user, and organization-user reports can expose the activity in one-day and 28-day windows. The new totals_by_3rd_party_agent data includes a stable agent_id and a display name that may change; the identifier should be the join key. This gives administrators a finer view of cost, permissions, and workflow adoption, but it does not automatically explain business value. Start with a read-only sample, reconcile time zones, pagination, and overlapping windows, then associate agent activity with AI credits, members, repositories, and permission changes before changing budgets or access.
Sources
FAQ
What is this ENHE AI article about?
Anthropic introduced Claude Fable 5.1 and Claude Mythos 5.1 on September 1. They share one base model but use different safeguard and access profiles. Fable is generally available and is estimated to cost 25% less for typical token workloads, with savings of up to about 45% for highly agentic workloads. Enterprise Frontier Safeguards will keep customer data in infrastructure controlled by the customer while providing misuse detection. Mythos is offered through trusted access programs for cybersecurity and life sciences. Anthropic also described software vulnerability discovery, protein binder design, and GPU kernel optimization examples. For enterprise teams, the launch makes model selection a joint decision about capability, cost, data residency, and risk controls.
Why is this AI update worth watching?
Fable 5.1 面向一般可用，Mythos 5.1 通过可信访问计划提供。 Fable 5.1 典型工作负载预计比 Fable 5 低约 25%，高代理工作负载最高约 45%。 Enterprise Frontier Safeguards 将数据存放在客户控制的云基础设施中。 官方称新网络安全护栏的误报减少约 60%，并允许识别漏洞但不开发利用。
What does it mean for everyday AI users?
ENHE 读者可把这次发布当作企业模型采购的评估模板：分别核算 token 成本、数据驻留、访问资格和护栏误报，而不是只比较榜单分数。
Where can readers continue learning on ENHE AI?
Readers can continue with ENHE AI software apps, AI skill tutorials, and AI account service guidance to turn the news signal into practical action.
