Claude Fable 5.1
Anthropic · released
Anthropic's top publicly available tier, sitting above Opus and aimed at the hardest reasoning problems and agent runs that last for hours. Anthropic itself suggests starting on Opus 5.5 and moving up only when that falls short — and the newer Opus now outscores it on most benchmarks.
Best for: Long-horizon agents · Hard reasoning · Deep research
- Intelligence Index
- 53.4
- #3 of 38 · at Max effort
- Price per 1M tokens
- $10 in $50 out
- Blended $20 at 3:1
- Context window
- 1M
- Up to 128K output
- Output speed
- 66 tokens/s
- $7.63 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall91 · #3 of 38
Based on 3 of 3 benchmarks
- Coding87 · #5 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work94 · #1 of 40
Based on 6 of 6 benchmarks
- Reasoning & science95 · #2 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability95 · #2 of 39
Based on 2 of 2 benchmarks
- Vision87 · #6 of 31
Based on 1 of 2 benchmarks
Reasoning effort trade-off
The same model at each effort setting. Lower settings give up a few points for a fraction of the cost and wait.
| Setting | Intelligence Index | Cost / task | First answer |
|---|---|---|---|
| Max | 53.4 | $7.63 | 4.9 min |
| xHigh | 53.2 | $5.98 | 81s |
| High | 51.2 | $3.91 | 23s |
| MediumSweet spot | 48.9 | $2.98 | 7.6s |
| Low | 46.8 | $2.37 | 3.9s |
Measured by Artificial Analysis. First answer is the median wait until the first answer token, including thinking time. “Sweet spot” is the cheapest setting that keeps at least 90% of the top score for no more than 60% of its cost.
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#3 of 38· Max effort· Artificial Analysis
- Vals Index65.8%
#4 of 26· Max effort· $28.7 per test· Vals AI
- Arena · Text1501 ±6
#3 of 33· Max effort· 11,800 votes· Arena
Coding
- Terminal-Bench 4.052.0%
#6 of 38· Max effort· Artificial Analysis
- SciCode63.1%
#2 of 38· Max effort· Artificial Analysis
- Arena · WebDev1745 ±9
#5 of 34· Max effort· 6,521 votes· Arena
- Vibe Code Bench90.3%
#5 of 26· Max effort· Vals AI
- Code Migration60.0%
#6 of 26· Max effort· Vals AI
Agents & knowledge work
- GDPval-AA62.9
#3 of 39· Max effort· Artificial Analysis
- Arena · Agent+14.3 ±1.9
#1 of 29· Max effort· Arena
- Finance Agent v258.9%
#4 of 26· Max effort· Vals AI
#1 of 26· Max effort· Vals AI
- Legal Research Bench55.3%
#1 of 26· Max effort· Vals AI
- Tax Agent Bench77.6%
#1 of 26· Max effort· Vals AI
Reasoning & science
- Humanity's Last Exam59.1%
#2 of 39· Max effort· Artificial Analysis
- CritPt29.7%
#6 of 39· Max effort· Artificial Analysis
- ARC-AGI-2saturating90.0%
#4 of 16· Max effort· $4.49 per task· ARC Prize
- ARC-AGI-3Not tested
- GPQA Diamondsaturating93.7%
#4 of 27· Max effort· Artificial Analysis
Knowledge & reliability
#2 of 39· Max effort· Artificial Analysis
- AA-LCR85.3%
#4 of 39· Max effort· Artificial Analysis
Vision
- MMMU-ProNot tested
- Arena · Vision1288 ±10
#5 of 19· Max effort· 4,312 votes· Arena
Reported by Anthropic
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- Terminal-Bench 4.0 · Max effort55.8%
- SWE-bench Pro · Max effort81.2%
- OSWorld 2.0 · partial score77.9%
- Humanity's Last Exam · no tools60.9%
Specifications
- Lab
- Anthropic
- Released
- Sep 1, 2026
- Availability
- Generally available
- Context window
- 1,000,000 tokens
- Max output
- 128,000 tokens
- Accepts
- Text, Images, PDFs
- Knowledge cutoff
- Jun 2026
- Weights
- Closed
- Reasoning control
- Always on · effort low to max
- API model ID
claude-fable-5-1
Pricing
- Input / 1M tokens
- $10
- Cached input / 1M
- $0.25
- Output / 1M tokens
- $50
- Blended (3:1)
- $20
- Measured cost / task
- $7.63
Full 1M context at standard rates. Batch API is half price.
Worth knowing
- Safeguards hand some cybersecurity and life-science requests to an older Opus model. Independent evaluators score what the product returns, so their results include those fallbacks.
- Claude Mythos 5.1 is the same model with looser safeguards, offered only to vetted security teams.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.