Claude 5.1 (Fable) in 2026: Anthropic's Mythos flagship — and why Opus 5 steals it
Anthropic's top Claude tier costs the same as OpenAI's flagship — then Anthropic released a half-price sibling that nearly matches it. The 5.1 story is really a two-model story.
- Mythos-tier flagship — Claude Fable 5.1 lists at $10 input / $50 output per million tokens, API-only — no subscription includes it.
- Ranked #3 for coding — Fable 5 holds #3 on the LLM Stats coding board (Sept 12, 2026) with 95.0% on SWE-bench Verified.
- The sibling steals it — Claude Opus 5 at $5/$25 posts near-flagship scores and ranks #3 of 232 models on BenchLM.
Anthropic shipped Claude Fable 5 on June 9, 2026 as the first general-use model of its Mythos class — the top tier above Opus. Since July 9 it is API-only: no subscription plan includes it, and every token bills at $10 per million input and $50 per million output with a flat-priced 1M-token context window.
Then came the plot twist. On July 24, Anthropic launched Claude Opus 5 at $5/$25 — half the price, near-Fable quality on most benchmarks. For most teams, the "which Claude" decision moved down a tier.
Fable 5.1 on the record
- Price: $10 input / $50 output per million tokens, API-only since July 9, 2026.
- Context: 1M tokens, flat pricing — no long-context surcharge step.
- Coding rank: #3 on the LLM Stats coding leaderboard (updated September 12, 2026) with a 45.7 rating, 95.0% on SWE-bench Verified, 84.3% on Terminal-Bench 2.1, and 80.0% on SWE-bench Pro.
- Peer below, close behind: Claude Mythos Preview sits at #4 with 93.9% on SWE-bench Verified.
Opus 5: the half-price problem
BenchLM's Opus 5 profile (data as of September 10, 2026) gives the sibling its due: a public score of 82/100, ranking #3 of 232 tracked models, with Agentic at #2, Coding at #3, Knowledge at #3, and Multimodal at #2. API pricing is $5 input / $25 output with cached input at just $0.50.
The cost-adjusted reading is blunt: Opus 5 lands within striking distance of the $10/$50 tier on most published rows at half the token price — and its cache economics ($0.50 cached input) dominate real bills on workloads with stable system prompts.
Against the field
Fable 5.1's $10/$50 headline matches GPT-6 Astra exactly — deliberate margin defense on both sides. Below them, Gemini 3.1 Pro prices at $2/$12 under 200K tokens ($4/$18 above), and newer GA options like Gemini 3.5 Pro with a 2M context keep the budget end moving. The premium tier is no longer anyone's default; it is a deliberate purchase for headroom you can name.
Buy Fable 5.1 ($10/$50) only when your task demonstrably needs the Mythos headroom — and prove it with a side-by-side on your own workload first, because the cheaper Claude usually wins the rematch.
How to decide in one afternoon
Run your hardest code patch and your longest document task through both models at your real input sizes. Score the outputs blind, then multiply by monthly volume at $5/$25 versus $10/$50. If Fable's margin doesn't survive that math, Opus 5 is your Claude.
Prices and benchmark figures are standard published rates as of September 12, 2026 and can change without notice. Some outbound links may be affiliate links — see our disclosure. Found an error? Tell us and we'll correct it in public.