Anthropic Claude Fable 5.1 Launches With Stronger Benchmarks as Samsung Tests Claude Code for Chip Design
Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, adding fresh momentum to the fast-moving AI race. The launch arrives ahead of OpenAI’s expected Astra-class models and gives Anthropic another opportunity to position Claude as a leading choice for advanced reasoning, coding, and technical workflows.
Claude Fable 5.1 and Mythos 5.1 are understood to be closely related models, with the key difference being that Fable 5.1 includes stronger safety guardrails. That makes Fable 5.1 the more restricted version, while Mythos 5.1 appears aimed at users who need fewer constraints within supported use cases.
The biggest highlight is performance. Claude Fable 5.1 reportedly delivers a notable improvement over Claude Fable 5 across several AI benchmarks. On the Artificial Analysis Intelligence Index, Fable 5.1 scores 66, marking a 4-point increase over Fable 5 and a 3-point lead over Opus 5, which previously scored 63.
The model also performs strongly in terminal-based testing. On Terminal-Bench-Science 0.1, Claude Fable 5.1 reaches 52.6%, more than doubling the result of Fable 5. On Terminal-Bench 4.0, it scores 55.8%, compared with 42.0% for Fable 5. These gains suggest that Anthropic has improved the model’s ability to handle technical, scientific, and command-line-style tasks.
Anthropic is also keeping API pricing mostly steady for Fable 5.1. The model remains priced at $10 per 1 million input tokens and $50 per 1 million output tokens, matching the earlier Fable 5.0 pricing. However, cache reads are now significantly cheaper, falling from $1 per 1 million tokens to $0.25 per 1 million tokens. That represents a 75% reduction and could make repeated or context-heavy workloads more affordable.
Still, the overall cost picture is more complicated. While cache reads are cheaper, the weighted average cost per Intelligence Index task is listed at $3.69 for Fable 5.1, compared with $2.34 for Fable 5.0. That means real-world complex tasks may cost around 58% more to run, likely because the newer model uses more compute to deliver stronger results.
Beyond benchmarks, Anthropic is gaining attention in enterprise AI through Claude Code, its coding-focused environment built around Claude models. Samsung’s LSI division is reportedly testing Claude for semiconductor design-related work, a major signal that AI tools are moving deeper into highly specialized engineering fields.
Choi Ki-young, CEO of Anthropic Korea, said Samsung Electronics is testing Claude in semiconductor design workflows. According to his comments, Samsung researchers have traditionally handled specialized chip design software manually, while AI automation in this area has remained limited.
Claude Code is reportedly being used to help generate verification environments, prepare test scenarios, and identify potential design issues. These are important steps in chip development, where even small mistakes can lead to costly delays. However, the AI’s work still requires human review, meaning Claude is being used as an assistant rather than a fully autonomous chip designer.
The Samsung testing is significant because semiconductor design is one of the most demanding technical fields for AI. It requires deep reasoning, code generation, error detection, and the ability to work within strict engineering constraints. If Claude Code proves useful in this environment, it could strengthen Anthropic’s case that its models are suitable for advanced enterprise and engineering applications.
The launch of Claude Fable 5.1 also shows how AI companies are now competing on multiple fronts at once: raw benchmark performance, API cost, coding capability, enterprise adoption, and specialized industry use cases. Anthropic’s latest model appears to make meaningful progress in performance while also lowering some usage costs, particularly around cache reads.
At the same time, the higher cost per complex benchmark task suggests that improved intelligence may still come with a compute trade-off. For businesses, the key question will be whether Claude Fable 5.1’s stronger performance justifies the additional cost in demanding workflows.
With Claude Fable 5.1, Mythos 5.1, and Claude Code gaining traction, Anthropic is clearly aiming beyond general-purpose chatbots. Its strategy increasingly targets technical users, developers, researchers, and enterprise teams that need AI capable of handling complex tasks. Samsung’s interest in using Claude for chip design only adds to the sense that AI is becoming a practical tool in some of the world’s most advanced engineering environments.
