# Claude 5 The Claude 5 generation is [[Anthropic]]'s 2026 model family. In this generation, two of the models share weights, one tier is not publicly available, and some queries sent to one model are answered by another. ## The lineup | Model | API string | Tier | Public | |---|---|---|---| | Claude Mythos 5 | — | Mythos | No | | [[Claude Fable 5]] | `claude-fable-5` | Mythos | Yes | | [[Claude Fable 5.1]] | `claude-fable-5-1` | Mythos | Yes | | Claude Mythos 5.1 | — | Mythos | No (verification programs only) | | [[Claude Opus 5]] | `claude-opus-5` | Opus | Yes | | [[Claude Opus 5.5]] | `claude-opus-5-5` | Opus | Yes | | [[Claude Sonnet 5]] | `claude-sonnet-5` | Sonnet | Yes | | [[Claude Sonnet 5.5]] | `claude-sonnet-5-5` | Sonnet | Yes | | [[Claude Haiku 5.5]] | `claude-haiku-5-5` | Haiku | Yes | | Claude Haiku 4.5 | `claude-haiku-4-5-20251001` | Haiku | Yes (previous generation) | [[Claude Opus 5.5]] (22 September 2026) opened the Claude 5.5 family, followed by [[Claude Sonnet 5.5]] (28 September 2026) and [[Claude Haiku 5.5]] (7 October 2026). Until then, Haiku was still on 4.5. All four current models (Fable 5.1 and the three 5.5 models) share a 1M-token context window, 128K max output and a June 2026 knowledge cutoff. Haiku 5.5 is the only one whose price depends on prompt length: five times more for prompts over 100K tokens. ## The Mythos tier Above Opus sits a new tier called **Mythos**. **Fable 5 and Mythos 5 are the same underlying model.** Fable carries additional safety measures for biology, cybersecurity, and LLM R&D. Mythos does not. The public frontier model and the restricted one are two safety configurations of one set of weights. **Claude Mythos Preview is not publicly available.** It runs with a small number of trusted organizations under **Project Glasswing** (anthropic.com/glasswing). The 5.1 point release kept the pattern: [[Claude Fable 5.1]] and Claude Mythos 5.1 are one set of weights, and Mythos 5.1 is only available through Anthropic's Cyber and Life Sciences Verification Programs. Until this generation, the tier list ordered models by capability. It now also encodes access policy: the top tier isn't available to the public. ## Safeguards routing When you select [[Claude Fable 5]], some queries get **silently redirected to [[Claude Opus 5]]**. Anthropic's own wording: > Without safeguards, Fable 5's capabilities in areas like cybersecurity could be misused to cause serious damage. We've therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 5. To release the model both safely and quickly, we've tuned these safeguards conservatively. They say it triggers in under 5% of sessions on average and admit it will sometimes catch harmless requests. So the model you selected is not always the model that answered. If you evaluate Fable 5 on a task that touches cybersecurity, some fraction of your results are Opus 5 results, and Fable 5 benchmark numbers on security-adjacent tasks need to account for this. The Opus 5 system prompt carries a dedicated `<fable_safeguards_routing>` block instructing the model on how to explain this to confused users. ## The export control interruption This happened after the models' training cutoff, so they have to be told about it: - **9 June 2026**: Fable 5 and Mythos 5 released - **12 June 2026**: Anthropic suspends access to both, to comply with U.S. Department of Commerce export controls - **30 June 2026**: The Department lifts the controls - **1 July 2026**: Access restored Access was suspended for about three weeks. See [[2026-06-17 The US Government Banned Claude Fable 5 and Mythos 5|the news note on the ban]]. ## What changed for the people using them **These models need less instruction than their predecessors.** Anthropic deleted over 80% of [[Claude Code]]'s system prompt for Opus 5 and Fable 5 with no measurable loss on coding evals. Their own transcripts showed conflicting guidance inside a single request ("leave documentation as appropriate" against "DO NOT add comments"), forcing the model to spend reasoning on reconciling its instructions rather than on the task. The six reversals are documented in [[Context Engineering]]. The short version: give judgment instead of rules, design interfaces instead of supplying examples, disclose progressively instead of loading upfront, stop repeating yourself, let auto-memory work, and use rich references instead of plain specs. Instructions written for the 4.x generation can work against these models. Run `/doctor` in [[Claude Code]]. ## Reasoning effort has a ceiling Vals AI measured [[Claude Opus 5]] across effort levels: performance rises from low to medium to high, then flattens or dips at the two highest levels while costing considerably more. [[GPT-5.6]]'s migration guidance (July 2026) also tells you to test one level below your current setting. For [[Claude Sonnet 5.5]] at Max effort, Artificial Analysis measured $7.60 per task (about 193k output tokens per task), more than Opus 5.5 at Max ($5.98) for a lower score. On FrontierCode, its Max score is also lower than its Xhigh score. ## Caveats - The tier structure and routing details here come substantially from a leaked Opus 5 system prompt, not from a formal product page. The behaviour is corroborated by Anthropic's own safeguards blog quote, but treat specifics as well-sourced rather than official - Model strings and tier names in this generation have moved fast. Verify against current docs before writing code against them - Mythos-tier availability may change ## References - [[Claude Opus 5]] announcement: https://www.anthropic.com/news/claude-opus-5 - Anthropic on Fable/Mythos access restoration: https://anthropic.com/news/fable-mythos-access - Project Glasswing: https://anthropic.com/glasswing - Thariq, "The new rules of context engineering for Claude 5 models" (2026-07-24, 4.3M views): https://x.com/trq212/status/2080710971228918066 - Anthropic blog version: https://claude.com/blog/the-new-rules-of-context-engineering-for-claude-5-generation-models - [[Hacker News]] discussion of the context engineering post (462 points): https://news.ycombinator.com/item?id=49051361 - Opus 5 system prompt leak, Pliny / CL4R1T4S (2026-07-25): https://x.com/elder_plinius/status/2080782804435251373 · https://github.com/elder-plinius/CL4R1T4S/blob/main/ANTHROPIC/OPUS-5.md - Vals AI on reasoning levels: https://x.com/ValsAI/status/2080817011157094511 - Current lineup comparison (Claude Haiku 5.5 overview, platform docs): https://platform.claude.com/docs/en/models/haiku-5-5/overview ## Related - [[Claude Opus 5]] - [[Claude Opus 5.5]] - [[Claude Fable 5]] - [[Claude Sonnet 5]] - [[Claude Sonnet 5.5]] - [[Claude Haiku 5.5]] - [[Claude Fable 5.1]] - [[Claude Opus 4.7]] - [[Anthropic]] - [[Claude]] - [[Claude Code]] - [[Context Engineering]] - [[GPT-5.6]] - [[AI Frontier Model]] - [[Claude Code effort levels]]