Anthropic released two new Claude AI models on September 1, with Fable 5.1 positioned as a cost-reduced upgrade for coding and knowledge work, and Mythos 5.1 remaining available only to approved users.

Fable 5.1 is a point-release iteration of the company's mid-tier coding model, lowering inference costs while maintaining or improving performance on programming tasks and information retrieval. Mythos 5.1, Anthropic's most capable reasoning model, stays behind an approval gate. The company did not disclose pricing for either release in the announcement.

AnthropicClaude models sit in a three-tier structure: Haiku for lightweight tasks, Fable for coding and knowledge work, and Mythos for complex reasoning. Point releases typically refine existing models without architectural changes. The company last released significant Claude updates in May 2025 with Claude 3.5 Sonnet.

The timing follows continued competition in AI model releases. OpenAI released GPT-4o in May 2024 and has maintained a cadence of incremental improvements. Google released Gemini 2.0 in December 2024 with reasoning capabilities. Both competitors offer tiered access models with premium tiers restricted to paid users or waitlists.

MSB Intel

Anthropic has emphasized cost reduction across its product line. Earlier releases of Claude 3.5 Sonnet included price cuts compared to prior generations. Lower inference costs on Fable 5.1 could expand adoption among developers and enterprises managing large token volumes for routine coding tasks.

Mythos 5.1 remaining gated means the company is managing capacity or conducting extended testing before wider release. Anthropic did not specify the approval criteria or timeline for broader access.

Fable 5.1's cost reduction on a model already targeted at coding workloads narrows the price gap with open-source alternatives like Llama, which developers can run locally. Anthropic has not disclosed whether the performance gains on Fable 5.1 offset the cost difference enough to shift developer preference away from smaller models.

The company to watch is Anthropic: whether Mythos 5.1 reaches general availability within 90 days will determine the urgency of competing on reasoning capability breadth versus OpenAI's phased access strategy.