Claude Fable 5.2 vs GPT-6 Astra
Claude Fable 5.2 costs $8.00 / $40.00 and GPT-6 Astra costs $10.00 / $50.00 per million input / output tokens. At a typical 3:1 mix, Claude Fable 5.2 is 1.3× cheaper than GPT-6 Astra. Claude Fable 5.2 also scores clearly higher on Epoch AI’s Capabilities Index (174 vs 169).
Price
Claude Fable 5.2 is 1.3× cheaper per token
On a coding agent: $1,521 vs $2,598 a month (41% less)
Capability
Claude Fable 5.2 scores higher
Claude Fable 5.2 174 (170–179)
GPT-6 Astra 169 (166–173)
GPT-6 Astra 169 (166–173)
Best value at Claude Fable 5.2’s level
Claude Fable 5.2
The cheapest current model that scores level or higher, on a coding agent
Input
$8.00
per 1M
Output
$40.00
per 1M
Cached
$0.20
1M context
Input
$10.00
per 1M
Output
$50.00
per 1M
Cached
$1.00
1.05M context
Side by side
✓ marks the cheaper price, larger context window, bigger discount or higher score.
| Detail | Claude Fable 5.2 | GPT-6 Astra |
|---|---|---|
| Price per 1M tokens | ||
| Input | $8.00Lowest | $10.00 |
| Output | $40.00Lowest | $50.00 |
| Blended (3 in : 1 out) | $16.00Lowest | $20.00 |
| Cached input | $0.20Lowest | $1.00 |
| Batch or off-peak discount | 50% (batch) | 50% (batch) |
| Long prompts | Same rate at any length | Over 272K: 2× in, 1.5× out |
| Limits and model | ||
| Context window | 1M | 1.05MLargest |
| Capability score Epoch AI index | 174Highest | 169 |
| Arena text score 25 Sep 2026 | — | 1478 |
| Released | 9 Nov 2026 | 3 Sep 2026 |
| Open weights | No | No |
| Status | Current | Current |
| Checked | ||
| Price checked | 11 Nov 2026 Anthropic pricing | 2 Nov 2026 OpenAI pricing |
Monthly cost on three workloads
List prices, no batch discount. The cached share applies where a model has a cache rate.
| Workload | Claude Fable 5.2 | GPT-6 Astra | Difference |
|---|---|---|---|
| Customer chat assistant50,000 conversations a month, about 2,000 tokens in and 400 out each; 50% cachedPrice this workload in the calculator | $1,210Lower | $1,550 | $340 |
| Document Q&A (RAG)20,000 questions a month over retrieved documents, 12,000 in and 600 out; 30% cachedPrice this workload in the calculator | $1,838Lower | $2,352 | $514 |
| Coding agent400 tasks a month, 40 steps each with a growing context; 85% cachedPrice this workload in the calculator | $1,348Lower | $2,163 | $815 |
How they differ
- Claude Fable 5.2 is cheaper on input ($8 vs $10).
- Claude Fable 5.2 is cheaper on output ($40 vs $50), which matters most for long answers and reasoning models.
- GPT-6 Astra accepts longer prompts (1.05M vs 1M tokens).
- Cached input is cheaper on Claude Fable 5.2 ($0.20 vs $1), which favours agents and chat apps that resend the same context.
- On Epoch AI’s Capabilities Index, Claude Fable 5.2 scores clearly higher (174 vs 169), beyond the margin of error.
Recent changes
All changes- ▼ 20%Claude Fable 5.2 replaced Fable 5.1 at $8 / $40 per 1M tokens, 20% less, with cache reads at $0.20. Fable 5.1 stays available at $10 / $50.
$10 / $50→ $8 / $40 - OpenAI launched GPT-6.1 Terra at $1 / $5 per 1M tokens, with cached input at $0.05: half the price of GPT-6.1 Sol.
- GPT-6.1 Sol replaced GPT-6 Sol at the same $2 / $10 per 1M tokens, with cached input cut from $0.20 to $0.10.
- OpenAI launched GPT-6 Sol at $2 / $10 and GPT-6 Luna at $0.10 / $0.50.
- OpenAI launched GPT-6 Astra at $10 / $50, its most expensive standard model.
