Claude Fable 5.1 — Anthropic's latest model with 75% cheaper cache reads
On September 1, 2026, Anthropic released Claude Fable 5.1 — a 1M-context, 128K-output model with dramatically cheaper cache reads and the same base rates as Fable 5. Here is the full pricing breakdown, what changed, and how to access it via API.
Claude Fable 5.1 — Anthropic's latest model with 75% cheaper cache reads
Claude Fable 5.1 is Anthropic's latest model, released September 1, 2026. It retains the 1M-token context window and 128K max output from Fable 5 but brings a major reduction in prompt-caching costs — 75% cheaper cache reads compared to Fable 5. The knowledge cutoff is June 2026.
Adaptive thinking is always on, with the default effort level set to high. The base input and output rates are the same as Fable 5, but the cache-read discount makes sustained agentic workloads significantly cheaper. Anthropic also released Claude Mythos 5.1 alongside it, a restricted-access variant for specific use cases.
For typical workloads, Fable 5.1 is roughly 25% cheaper than Fable 5, and for agentic workloads with heavy cache reuse, the savings can reach 45%.
Claude Fable 5.1 costs $10 per million input tokens and $50 per million output tokens — identical to Fable 5. The difference is in caching: cache reads drop from $1.00/MTok on Fable 5 to $0.25/MTok on Fable 5. Batch API users get 50% off the base rates ($5 input, $25 output). US-only inference carries a 1.1x multiplier.
| Model | Input | Cache Read | Cache Write (5min) | Cache Write (1hr) | Output |
|---|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $0.25 | $12.50 | $20.00 | $50.00 |
| Claude Fable 5.1 (Batch) | $5.00 | $0.25 | $12.50 | $20.00 | $25.00 |
Source: Anthropic official pricing as of September 1, 2026. US-only inference carries a 1.1x multiplier. Batch API applies 50% discount to base input/output rates.
The base rates for Fable 5.1 are identical to Fable 5 ($10/$50 per MTok), but cache reads are 75% cheaper ($0.25 vs $1.00). For a typical workload with 70% cache hits, this translates to roughly 25% overall savings. For agentic workloads with 90%+ cache hits, savings reach 45%.
Compared to Opus 5 ($5/$25), Fable 5.1's base rates are 2x higher, but cache reads are actually cheaper ($0.25 vs $0.50). This makes Fable 5.1 more cost-effective than Opus 5 for long-running agents with heavy prompt reuse, despite the higher sticker price on uncached tokens.
| Model | Input | Cache Read | Output | Context |
|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $0.25 | $50.00 | 1,048,576 |
| Claude Fable 5 | $10.00 | $1.00 | $50.00 | 1,048,576 |
| Claude Opus 5 | $5.00 | $0.50 | $25.00 | 1,048,576 |
Note: Cache reads on Fable 5.1 are cheaper than both Fable 5 and Opus 5, making it the most cost-effective choice for cache-heavy agentic workloads.
Fable 5.1 is a cost-reduction release, not a capabilities leap. If your workload already runs on Fable 5, upgrading to 5.1 is essentially free savings — same model quality, dramatically cheaper cache reads. For agentic applications with large, stable prompt prefixes, the 75% cache-read discount can cut total API spend by nearly half.
The main caveat: if your workload has low cache-hit rates (under 50%), the savings are modest, and the higher base rates compared to Opus 5 may not justify the switch. The sweet spot is long-running coding agents, multi-turn research sessions, and any application that repeatedly references a large codebase or document set.
APItokendeal provides OpenAI-compatible access to supported Claude, GPT, Gemini, and other models through one key. Teams can evaluate Fable 5.1 against other model families without maintaining a separate integration for each provider.
APItokendeal provides OpenAI-compatible API access to a wide range of models including Claude Fable 5.1, GPT, Gemini, DeepSeek, and more. One API key, one dashboard, prepaid token packs with no surprise bills.