Anthropic
Claude Fable 5.2
Anthropic's most capable generally available model. It lists 20% below Fable 5.1, generates up to 30% faster and leads most of Anthropic's launch table.
Replaces Claude Fable 5.1: input $10.5 → $8.4 per 1M (-20%), output $52.5 → $42 per 1M (-20%), same context window.
Run this model
anthropic/claude-fable-5.2
Use an endpoint supported by this model.
Get a key for this setup →const response = await fetch('https://api.minirouter.sh/v1/chat/completions', { method: 'POST', headers: { Authorization: `Bearer ${process.env.MINIROUTER_KEY}`, 'Content-Type': 'application/json', }, body: JSON.stringify({ model: 'anthropic/claude-fable-5.2', messages: [{ role: 'user', content: 'Why is the sky blue?' }], max_tokens: 1024, }),}) const data = await response.json()console.log(data.choices[0].message.content)Thinking is always on and uses output tokens, so leave room for the answer in max_tokens. Sampling parameters have no effect.
Model overview
- Input / 1M
- $8.4
- All-in MiniRouter rate
- Output / 1M
- $42
- All-in MiniRouter rate
- Context
- 1M
- 1,000,000 tokens
- Max output
- 128K
- tokens
- Effort
- low → max
- adaptive thinking, always on
- Speed
- Up to 30% faster
- than Fable 5.1, per Anthropic
- Announced
- 9 Nov 2026
- publisher release
Compare text-model alternatives or estimate token costs for your workload.
Ahead on coding, research and long tasks.
Anthropic's launch table puts Fable 5.2 first on ten of these twelve evaluations. Gemini 4 Carbon leads AutomationBench and Opus 5.5 leads the AA-Briefcase Elo rating. MiniRouter plotted the table and did not run the evaluations.
Agentic coding
The top FrontierCode score among released models Anthropic tested.
Real engineering tasks. At High effort Fable 5.2 scores 45.3%, level with Opus 5.5's best, for roughly half the cost per task.
- Fable 5.2
- 49.1%Xhigh effort
- Fable 5.1
- 32.4%Xhigh effort
- Opus 5.5
- 45.5%Max effort
FrontierCode 2.0
0 to 100%
- Claude Fable 5.249.1%1
- Claude Fable 5.132.4%
- Claude Opus 5.545.5%
- GPT-6 Astra42.8%
- Gemini 4 Carbon41.1%
View chart data
| Model | Score |
|---|---|
| Claude Fable 5.2 | 49.1% |
| Claude Fable 5.1 | 32.4% |
| Claude Opus 5.5 | 45.5% |
| GPT-6 Astra | 42.8% |
| Gemini 4 Carbon | 41.1% |
Eleven more evaluations
Coding, research, computer use, knowledge work
- Claude Fable 5.2
- Comparison models
Agentic coding
FrontierSWE v2
- Claude Fable 5.269.9%
- Claude Fable 5.156.3%
- Claude Opus 5.562.3%
- GPT-6 Astra65.5%
- Gemini 4 Carbon61.2%
0 to 100%
Business workflows
AutomationBench
- Claude Fable 5.251.4%
- Claude Fable 5.131.4%
- Claude Opus 5.542.5%
- GPT-6 Astra41.4%
- Gemini 4 Carbon53.9%
0 to 100%
Agentic scientific research
Terminal-Bench-Science 0.1
- Claude Fable 5.274.6%
- Claude Fable 5.152.6%
- Claude Opus 5.558.7%
- GPT-6 Astra68.1%
- Gemini 4 Carbon64.8%
0 to 100%
Agentic search
WANDR · relaxed score
- Claude Fable 5.282.1%
- Claude Fable 5.168.7%
- Claude Opus 5.572.3%
- GPT-6 Astra—Not reported
- Gemini 4 Carbon—Not reported
0 to 100%
Computer use
OSWorld 2.1 · offline set, partial
- Claude Fable 5.289.3%
- Claude Fable 5.183.5%
- Claude Opus 5.587.2%
- GPT-6 Astra79.5%
- Gemini 4 Carbon81.5%
0 to 100%
Robotics
RoboDojo Sim
- Claude Fable 5.264.9%2
- Claude Fable 5.127.6%
- Claude Opus 5.539.4%
- GPT-6 Astra29.8%
- Gemini 4 Carbon—Not reported
0 to 100%
Knowledge work
AA-Briefcase v1.1 · rubric
- Claude Fable 5.266.1%
- Claude Fable 5.161.5%
- Claude Opus 5.561.2%
- GPT-6 Astra52.0%
- Gemini 4 Carbon61.8%
0 to 100%
Knowledge work
AA-Briefcase v1.1 · Elo
- Claude Fable 5.21,785
- Claude Fable 5.11,669
- Claude Opus 5.51,800
- GPT-6 Astra1,569
- Gemini 4 Carbon1,611
1,500 to 1,900 Elo
Knowledge work
Remote Labor Index
- Claude Fable 5.241.3%3
- Claude Fable 5.117.9%
- Claude Opus 5.521.3%
- GPT-6 Astra20.8%
- Gemini 4 Carbon—Not reported
0 to 100%
Multidisciplinary reasoning
Humanity's Last Exam · with tools
- Claude Fable 5.274.2%
- Claude Fable 5.165.6%
- Claude Opus 5.567.7%
- GPT-6 Astra57.2%
- Gemini 4 Carbon—Not reported
0 to 100%
Visual chart recognition
Chartography · no tools
- Claude Fable 5.273.4%
- Claude Fable 5.144.8%
- Claude Opus 5.564.4%
- GPT-6 Astra71.0%
- Gemini 4 Carbon72.5%
0 to 100%
View chart data
| Benchmark | Claude Fable 5.2 | Claude Fable 5.1 | Claude Opus 5.5 | GPT-6 Astra | Gemini 4 Carbon |
|---|---|---|---|---|---|
| FrontierCode 2.0 | 49.1% | 32.4% | 45.5% | 42.8% | 41.1% |
| FrontierSWE v2 | 69.9% | 56.3% | 62.3% | 65.5% | 61.2% |
| AutomationBench | 51.4% | 31.4% | 42.5% | 41.4% | 53.9% |
| Terminal-Bench-Science 0.1 | 74.6% | 52.6% | 58.7% | 68.1% | 64.8% |
| WANDR · relaxed score | 82.1% | 68.7% | 72.3% | Not reported | Not reported |
| OSWorld 2.1 · offline set, partial | 89.3% | 83.5% | 87.2% | 79.5% | 81.5% |
| RoboDojo Sim | 64.9% | 27.6% | 39.4% | 29.8% | Not reported |
| AA-Briefcase v1.1 · rubric | 66.1% | 61.5% | 61.2% | 52.0% | 61.8% |
| AA-Briefcase v1.1 · Elo | 1,785 | 1,669 | 1,800 | 1,569 | 1,611 |
| Remote Labor Index | 41.3% | 17.9% | 21.3% | 20.8% | Not reported |
| Humanity's Last Exam · with tools | 74.2% | 65.6% | 67.7% | 57.2% | Not reported |
| Chartography · no tools | 73.4% | 44.8% | 64.4% | 71.0% | 72.5% |
Every list rate is 20% lower.
Anthropic list, per 1M tokens
- Input
- $8.00was $10.00
- Output
- $40.00was $50.00
- Cache write · 5 min
- $10.00was $12.50
- Cache read
- $0.20was $0.25
Output list price
USD per 1M output tokens
$0 to $50
- Claude Fable 5.240
- Claude Fable 5.150
View chart data
| Rate | Claude Fable 5.2 | Claude Fable 5.1 |
|---|---|---|
| Input / 1M | $8.00 | $10.00 |
| Output / 1M | $40.00 | $50.00 |
| Cache write, 5 min / 1M | $10.00 | $12.50 |
| Cache write, 1 hour / 1M | $16.00 | $20.00 |
| Cache read / 1M | $0.20 | $0.25 |
Four behaviours to plan for.
- [ 01 ]
Sampling parameters are ignored
temperature, top_p and top_k have no effect on the answer.
- [ 02 ]
Thinking cannot be turned off
Adaptive thinking is always on. Steer it with effort from low to max; the API default is high.
- [ 03 ]
Forced tool use is not enforced
A required or named tool_choice behaves like "auto", so the model can answer in text. Name the tool in the prompt, or use structured outputs for JSON.
- [ 04 ]
Thinking blocks are bound to the model
Fable 5.1 cannot read Fable 5.2 thinking blocks, and edited history before a thinking block returns 400. Keep harnesses append-only.
One route, standard clients.
https://api.minirouter.sh/v1anthropic/claude-fable-5.2/v1/chat/completions/v1/messages/v1/responsesConfigured providers. Priority follows the live catalog order.
| Provider | Priority | Input / 1M | Output / 1M | Parameters | Access |
|---|---|---|---|---|---|
vercel | 1 | $8.4 | $42 | max_tokens, stop, tools, tool_choice, reasoning, include_reasoning | MiniRouter balance |