Claude Fable 5.1 vs Opus 5 vs Opus 5.5: Which Wins?

Anthropic shipped Claude Fable 5.1 on 1 September 2026, six weeks after Claude Opus 5 landed on 24 July. The two sit at the top of the Claude lineup with identical context windows and identical output limits — and a price gap of exactly 2×.

That 2× is where most comparisons stop. It's also the least interesting number here, because Fable 5.1 inverts one line of the price sheet: it charges half what Opus 5 charges for prompt-cache reads. For a long-running agent re-reading a big cached prefix every turn, that single line changes the answer completely. Below is the arithmetic.

One thing has changed since this comparison was first published, and you need it before the arithmetic. On 22 September 2026 Anthropic released Claude Opus 5.5 at $4/$20 per million tokens and moved Claude Opus 5 into the "Legacy models (still available)" list in its models overview. Opus 5 still serves requests at its published prices and the head-to-head below is still accurate — but if you are picking a model today rather than auditing a deployment you already run, Opus 5.5 is the model to compare against. It also undercuts Fable 5.1 on cache reads, which removes Fable 5.1's one pricing advantage. Both comparisons are below: Opus 5 because it is what the question asks, Opus 5.5 because it is what you should actually buy.

How do Claude Fable 5.1 and Opus 5 compare on specs?

Both models are pinned snapshots on the Claude API — claude-fable-5-1 and claude-opus-5. Everything below is from Anthropic's models overview and pricing page.

Three rows deserve a second look. The cache-read row is the only place Fable 5.1 undercuts Opus 5, and it does so by half. Opus 5's fast mode bills at $10/$50 — exactly Fable 5.1's list price — for up to 2.5× the output speed, so if you have budget for Fable 5.1 rates, a faster Opus 5 is a real alternative rather than a consolation prize.

The third row is the one that ends the argument. Opus 5.5's cache reads are $0.20/MTok — below Fable 5.1's $0.25. Anthropic introduced a third cache multiplier for it (0.05x base input, against 0.025x on Fable 5.1 and the standard 0.1x everywhere else), and because Opus 5.5's base input is $4 rather than $10, the lower multiplier on the pricier model loses. Fable 5.1 keeps the lowest cache multiplier in the lineup and no longer has the lowest cache price. Every "Fable 5.1 gets cheaper the more you cache" argument below applies to Opus 5 only.

What do the benchmarks actually say?

Anthropic published a same-harness comparison across three models in its Fable 5.1 announcement. These are vendor-reported numbers — Anthropic ran them — but they at least put all three models on the same evals.

That table is the whole argument. Fable 5.1 wins all five — but outside one benchmark, by 1.4 to 3.5 points. The single blowout is Terminal-Bench-Science, where it nearly doubles Opus 5 (52.6% vs 29.0%) — a scientific-research agentic eval, not a general coding one. That is what Anthropic means in positioning Fable 5.1 as a research and long-horizon model rather than a general upgrade.

Independent measurement agrees on the size of the gap, with one large caveat about which version of each model you are reading. Artificial Analysis rebased its Intelligence Index from v4.1.1 to v4.3.2, and because a rebase changes the benchmark basket there is no conversion factor between old and new scores. These are v4.3.2 figures captured 5 October 2026, with every effort variant AA publishes for the two models:

Fable 5.1's win is a Max-effort win, and only that. Compared Max to Max it leads 53.35 to 50.78 — 2.57 points, consistent with the non-science benchmarks above. Compared at each model's default effort the ranking flips: Opus 5 scores 49.68 against Fable 5.1's 46.82. That matters because people quote these numbers across tiers without noticing. "Fable 5.1 beats Opus 5 on Artificial Analysis" is true at Max and false at default, and a comparison of unqualified Fable 5.1 against Opus 5 at Max is comparing unlike things.

The speed trade-off is the same story read sideways. At Max, Fable 5.1 streams 61.44 output tokens per second but takes a median 259 seconds to produce its first chunk, against 59 seconds for Opus 5 at Max and 33 seconds for Opus 5 at default. Drop Fable 5.1 to default effort and it answers in 3.4 seconds — and gives up 6.53 Index points. Always-on adaptive thinking has to happen somewhere, and it happens before you see anything.

And the row at the top of that table is the one to act on. Opus 5.5 scores 57.62, 4.27 points above Fable 5.1 at Max and 6.84 above Opus 5 at Max, at $5.98 per completed Index task against Fable 5.1's $7.63 — 21.6% less for the stronger model. One nuance worth keeping honest: against Opus 5 ($5.86/task) Opus 5.5 is about 2% more expensive per finished task despite charging 20% less per token, because it spends more tokens getting there. Lower rate card, slightly higher bill. Compare per task, not per token.

On human preference, LMArena (now served at arena.ai) on 5 October 2026 ranks claude-fable-5.1-max 6th at 1501±6 from 11,800 votes, claude-opus-5-high 13th at 1490±4 from 59,482 votes, and claude-opus-5.5-high 4th at 1504±9 from 4,552 votes. Two reads worth having. Fable 5.1's 11-point lead over Opus 5 is real but modest, and Opus 5.5 edges ahead of Fable 5.1 on human preference too — though with ±9 and ±6 intervals overlapping, that one is not yet a clean separation. Note also that claude-opus-5-high (1490) outscores claude-opus-5-max (1489): on open-ended chat, more thinking is not automatically better, which is the opposite of what the Intelligence Index shows for the same pair.

One honest gap: neither model's SWE-bench Verified score appeared in the launch materials or documentation we could verify. Anthropic used Terminal-Bench 4.0 and CursorBench 3.2.0 as its coding-agent evals instead, so those are what we've quoted.

Why does Fable 5.1 cost 2× on list but half on cache reads?

Anthropic's pricing page carries a footnote that does more work than the whole rest of the table: "Cache hits and refreshes on Claude Fable 5.1 and Claude Mythos 5.1 are priced at 0.025x the base input price. All other models use the standard 0.1x multiplier."

The mechanism matters more than the two dollar figures. Fable 5.1 sits on a 2.5% cache multiplier; Opus 5 — like Fable 5 and most of the catalogue — sits on the standard 10%. Because Fable 5.1's base input is $10 and Opus 5's is $5, that four-fold multiplier advantage more than cancels the two-fold base-price disadvantage, and the arithmetic lands on $0.25 against $0.50. Against Opus 5, Fable 5.1 is half price on the one token category that dominates agentic workloads. It's also a 4× cut from Fable 5's own $1.00/MTok.

That footnote now has a sibling. Anthropic added a 5% multiplier for Opus 5.5, and 5% of a $4 base is $0.20/MTok. So the generalisation to avoid is "the lowest multiplier wins" — what you pay is multiplier × base price, and Opus 5.5's higher multiplier on a much lower base beats Fable 5.1 outright. Fable 5.1's cache-read advantage is specific to the models it was launched against.

What does a stateless request cost on each?

A one-shot call with no caching — 10,000 input tokens, 2,000 output tokens:

• Fable 5.1: (10,000 × $10 + 2,000 × $50) ÷ 1M = $0.20
• Opus 5: (10,000 × $5 + 2,000 × $25) ÷ 1M = $0.10

Clean 2×. If your workload looks like this — stateless classification, one-off generation, RAG with a fresh context each time — the price gap is exactly what the sticker says and there is no clever argument to be had.

What does a cached coding agent actually cost?

Now a realistic long-running agent: a 200,000-token cached prefix (system prompt, tool definitions, repository context), 50 turns, each turn adding 2,000 new input tokens and producing 1,500 output tokens, using the 5-minute cache.

Against Opus 5 the gap has collapsed from 2× to 1.13×. Fable 5.1 costs 13% more, not 100% more. The cache-read line alone hands back $2.50, which nearly cancels the $1.88 output premium and the $1.25 cache-write premium combined. For a 2.57-point intelligence gain, a 13% premium is a genuinely different decision from a 100% one — and this is the strongest version of the case for Fable 5.1.

Against Opus 5.5 it does not survive. The same session costs $4.90, roughly half Fable 5.1's $9.75, because Opus 5.5 is cheaper on all four line items including the cache reads that were supposed to be Fable 5.1's edge. One caution on reading this table: it holds token counts identical across the three models, and they are not identical in practice — Opus 5.5 is measurably more verbose. The per-completed-task figures in the benchmark section are the honest totals, and they still favour Opus 5.5 over Fable 5.1 by 21.6%.

When does Fable 5.1 become genuinely cheaper?

Push the same logic further. Take a workload dominated by cache reads: a 1M-token cached document corpus held in the 1-hour cache, queried 500 times, each query adding 100 input tokens and producing 200 output tokens.

Against Opus 5, Fable 5.1 is 43% cheaper — a $112 saving on a $263 job, while also being the stronger model. That inversion is real, not a rounding artefact, and it is the single best argument for Fable 5.1 anywhere on this page.

Against Opus 5.5 the same workload reverses again. Opus 5.5 runs it for $110.20 against Fable 5.1's $150.50 — Fable 5.1 is 37% more expensive, on the exact workload shape that was supposed to be its home ground, and it scores 4.27 Index points lower. There is no token mix at which Fable 5.1 is the cheaper of those two, because Opus 5.5 is lower on every per-token line.

The general break-even against Opus 5, in tokens: Fable 5.1 wins on cost when cache_reads > (20 × new_input) + (25 × cache_writes) + (100 × output). The practical version is simpler — your cache-read volume has to exceed roughly 100× your output volume before the cheaper reads outweigh the doubled output price. Chatty agents that write a lot of code never get there. Retrieval and analysis loops over a large fixed corpus get there easily. Against Opus 5.5, there is no break-even to compute.

One caveat on Anthropic's own framing. The announcement says costs drop "approximately 25% for typical workloads and up to 45% for highly agentic tasks" — that is Fable 5.1 versus Fable 5, not versus Opus 5. It is easy to misread as a claim about this matchup. It isn't one.

What about batch and fast mode?

The Batch API takes 50% off input and output on both models, and stacks with cache multipliers. That produces a result worth knowing:

• Fable 5.1 on batch is $5 / $25 — identical to Opus 5 at list price. If your workload tolerates asynchronous processing, you can run the stronger model for what the cheaper model costs synchronously.
• Opus 5 on batch drops to $2.50 / $12.50, so the 2× ratio holds if you batch both. Opus 5.5 on batch is $2 / $10 — the cheapest way to run any current frontier Claude.
• Fable 5.1 does not support fast mode; both Opus models do. Opus 5's fast mode is $10 / $50, exactly Fable 5.1's list price, for up to 2.5× output speed. Opus 5.5's is $8 / $40. Either is the right trade when latency, not intelligence, is your constraint.
• Fable 5.1 is also the only one of the three without the 300K-output batch beta (output-300k-2026-03-24). If a single response has to exceed 128K tokens, Fable 5.1 cannot do the job at any price.

Which model is better for coding and agentic work?

Between those two, Fable 5.1 — but by less than the price implies. It leads Terminal-Bench 4.0 55.8% to 52.3% and CursorBench 3.2.0 73.4% to 70.0%. Anthropic's own Opus 5 announcement claimed Opus 5 lands "within 0.5% of Fable 5's peak score" on CursorBench 3.2 at maximum effort "but at half the cost per task" — and Fable 5.1 has only moved that needle 3.4 points.

Across all three, Opus 5.5 leads both, and it is not close on the agentic-coding evals. Anthropic's Opus 5.5 announcement publishes a same-harness three-way table in which Opus 5.5 wins every row:

These are a later benchmark generation than the Fable 5.1 launch table above — CursorBench 4.0 not 3.2.0, OSWorld 2.1 not 2.0 — so don't read figures across the two tables. Within this one, Opus 5.5's margin over Fable 5.1 runs from 0.6 points on Chartography to 10.6 on Terminal-Bench 4.0, and Anthropic's own summary is that Opus 5.5 "performs at the level of Claude Fable 5.1 on most work." The published rows put it a little above that.

Where Fable 5.1 pulls decisively ahead is duration. Anthropic lists its gains as concentrated in agentic coding over sessions that run for hours, multistep research that follows up on what it finds, long-context reasoning across the full 1M window, and computer use with recovery from failed steps. If your agent runs for twenty minutes, Opus 5 is fine. If it runs for four hours across a large refactor, the compounding error rate is what you're paying to reduce.

Two countervailing details sit in Anthropic's migration notes. Fable 5.1's parallel tool calling is more variable — it may issue one tool call per turn where Fable 5 batched several, costing round trips and wall-clock time. It is also "more likely to rewrite the entire file than make a targeted edit," burning output tokens at $50/MTok. Both are prompt-fixable, but neither shows up in a benchmark score and both work against the cost model above.

Where is each model genuinely worse?

Fable 5.1's real drawbacks:

• Forced tool use returns a 400 error. tool_choice of any or tool is rejected outright: tool_choice: type "tool" and "any" are not supported for this model. The cause is structural rather than an oversight — thinking is always on, and a forced call would skip it, pushing the model's working-out into the tool arguments and degrading their quality. auto and none still work, and Anthropic's suggested replacements are strict tool use or structured outputs. If your integration depends on pinned tool selection, migrating is code work, not a string swap. Note this is no longer a Fable-only constraint: Opus 5.5 rejects forced tool use too, so staying on Opus means staying on legacy Opus 5.
• Thinking cannot be disabled. thinking: {"type": "disabled"} errors. You pay for reasoning tokens at $50/MTok on every call, including trivial ones. Opus 5.5 shares this restriction; Opus 5 does not.
• Editing earlier turns invalidates thinking blocks. Thinking blocks are bound to both the model that produced them and the conversation prefix, so changing the system prompt, the tools array, or any earlier message errors the next request. Harnesses that rebuild the messages array must become append-only. The check is enforced for accounts created on or after 31 August 2026.
• Latency. Anthropic labels it "Slower" against "Moderate" for both Opus models; Artificial Analysis v4.3.2 puts median time-to-first-token at 259 seconds at Max effort, against 59 seconds for Opus 5 at Max.
• Data retention. Fable 5.1 carries 30-day retention and is not available under zero data retention unless Anthropic expressly authorises it. For some regulated buyers that ends the conversation before price does.
• No fast mode. At low effort it answers from memory more often instead of calling search tools.

Opus 5's real drawbacks:

• It is a legacy model. Anthropic's models overview moved Opus 5 into "Legacy models (still available)" when Opus 5.5 shipped on 22 September 2026, and its migration guide now points Opus 5 users at Opus 5.5. It keeps serving requests with a retirement commitment of not sooner than 24 July 2027, but it is no longer the model anyone should start a new integration on. That is the strongest argument against choosing it today, and it has nothing to do with benchmarks.
• A lower ceiling on the hardest reasoning. Terminal-Bench-Science 29.0% vs 52.6% is not a rounding difference — on novel scientific work, Opus 5 is not in the same class.
• Cache reads cost double. $0.50 vs $0.25/MTok is the one line where the cheaper model is the expensive one, and it dominates long agent loops.
• It trails on every published head-to-head benchmark, and its knowledge cutoff is a month earlier (May vs June 2026).

Which should you actually pick?

Anthropic's own guidance is unusually blunt for a vendor selling the more expensive model, and it has been re-pointed since Opus 5.5 shipped. Its Choosing a model page now states plainly: "Most workloads start with Claude Opus 5.5." The escalation rule is to move to Fable 5.1 only "if your evals at xhigh or max effort still fall short on demanding reasoning or long-horizon agentic work." The same page still calls Fable 5.1 "Anthropic's most capable model open to all customers" — a claim about long-horizon capability that its own benchmark table does not support on the scoreboard. Read it as positioning, and trust the evals you run.

That is the decision rule, and the benchmark spread supports it. Tuning the effort parameter is usually a better lever than switching models, and it's free.

One architectural gate sits upstream of any cost calculation. If your agent pins tool selection with tool_choice: {"type": "any"} or a named tool, Fable 5.1 is not a drop-in. Router layers that force a classification call, extraction pipelines mandating a schema tool, and state machines requiring a call at a known step all break with a 400 — as does any harness that mutates conversation history between turns. No benchmark or price table surfaces that. Check it before you model the tokens.

FAQ

Is Fable 5.1 better than Opus 5?

Yes, on every published head-to-head benchmark — but usually narrowly, and only at matched effort. Fable 5.1 leads Terminal-Bench 4.0 55.8% to 52.3%, CursorBench 3.2.0 73.4% to 70.0%, and scores 53.35 to 50.78 on Artificial Analysis's Intelligence Index v4.3.2 comparing each model's Max variant. At each model's default effort the order reverses: Opus 5 scores 49.68 to Fable 5.1's 46.82. The one decisive gap is Terminal-Bench-Science, 52.6% versus 29.0%. Outside scientific and long-horizon research work, expect two to three points, not a generational leap.

Is Opus 5.5 better than Fable 5.1?

Yes, on every number either company publishes. Anthropic's Opus 5.5 announcement has it ahead of Fable 5.1 on all nine benchmarks it lists, by 0.6 points (Chartography) to 10.6 points (Terminal-Bench 4.0). Artificial Analysis v4.3.2 scores Opus 5.5 at 57.62 against Fable 5.1's 53.35 at Max, and LMArena ranks claude-opus-5.5-high 4th at 1504 against claude-fable-5.1-max 6th at 1501. Opus 5.5 also costs $4/$20 against $10/$50, and $5.98 per completed Index task against $7.63. Anthropic's docs still describe Fable 5.1 as its most capable model open to all customers, but tell you to start with Opus 5.5.

Is Fable 5.1 worth 2× the price of Opus 5?

For most workloads, no — Anthropic itself says to start with Opus 5.5 now. Against Opus 5 the 2× only applies to uncached calls: in a realistic cached agent loop the premium falls to about 13%, and in cache-read-dominated workloads Fable 5.1 is outright cheaper. Against Opus 5.5 none of that holds, because Opus 5.5 is lower on every per-token line including cache reads. Work out your own token mix before accepting any sticker ratio as your actual cost, and compare cost per completed task rather than cost per million tokens.

Which is better for coding?

Fable 5.1, by 3.5 points on Terminal-Bench 4.0 and 3.4 on CursorBench 3.2.0. For most coding, that gap is not worth double the list price — raise Opus 5's effort parameter first. Fable 5.1 earns its keep on multi-hour autonomous sessions, large refactors, and migrations where compounding errors cost more than tokens.

Which is cheaper with prompt caching?

Against Opus 5, Fable 5.1 — it charges $0.25/MTok for cache reads against Opus 5's $0.50, 2.5% of base input versus the standard 10%. In a 200k-prefix, 50-turn agent loop that cuts Fable 5.1's premium from 100% to about 13%, and in a cache-read-dominated workload (roughly 100× more cache reads than output tokens) Fable 5.1 becomes 40%+ cheaper outright. Against Opus 5.5, neither: Opus 5.5's cache reads are $0.20/MTok — a 5% multiplier on a $4 base beats a 2.5% multiplier on a $10 base — so it is the cheapest of the three on every caching line.

What's the context window on each?

Both offer a 1M-token context window with 128K max output tokens, and both bill the full window at standard per-token rates — a 900k-token request costs the same per token as a 9k one. Both use the tokenizer introduced with Opus 4.7, so 1M tokens is roughly 555,000 words. There is no context-window advantage either way.

Should I switch from Opus 5 to Fable 5.1?

Probably switch to Opus 5.5 instead. Opus 5 is now a legacy model and Anthropic's migration guide points its users at Opus 5.5, which is stronger than both and cheaper per token than either. Move to Fable 5.1 only if your evals on Opus 5.5 at xhigh or max effort still fall short. Budget migration work for either destination: both reject forced tool use with a 400 error, neither lets you disable thinking, and on both, editing earlier turns invalidates thinking blocks — so any harness that rebuilds its message history needs changing first. Opus 5 is the only one of the three that still accepts tool_choice of type any or tool.

The bottom line

On the question as asked: Fable 5.1 is the better model and Opus 5 is the better default. A 2.57-point Index gap at Max effort does not justify a 2× bill for stateless work, and Anthropic's own documentation said as much. What genuinely changed that calculation was the cache-read inversion — at $0.25 against $0.50/MTok the expensive model gets cheaper the more you cache, and past roughly 100× cache reads to output tokens it won on both quality and cost.

On the question you should actually be asking: neither. Opus 5 went legacy on 22 September 2026, and Opus 5.5 beats Fable 5.1 on all nine benchmarks Anthropic published, on Artificial Analysis v4.3.2 (57.62 to 53.35), on cost per completed task ($5.98 to $7.63), and on cache reads ($0.20 to $0.25) — the one line that made Fable 5.1 defensible. The inversion that was Fable 5.1's whole argument does not exist against the current Opus.

So the decision rule is short. Start on Opus 5.5. Tune effort before you tune models, because it is free and usually the bigger lever. Escalate to Fable 5.1 only when your own evals at xhigh or max still fall short on hours-long autonomous work — which is the one case Anthropic reserves for it. And if you are still on Opus 5 today, migrate for the lifecycle status, not for a cost saving: per completed task, Opus 5.5 costs about 2% more than Opus 5, because the lower token rate is offset by heavier token use. Measure per job, never per million tokens.

Related reading: our Claude Opus 5.5 complete guide, the complete guide to Claude Fable 5.1, the earlier Opus 5 vs Fable 5 comparison, the Claude Opus 5 launch guide, and how Fable 5 usage credits work.