Find your next model
Choose your task. Compare the quality, estimate the cost, and try a model.
Models for your workload
249 matching models · 34 with a matched benchmark
| Model | Est. / month |
|---|---|
| $210.00 | |
| $210.00 | |
| $105.00 | |
| $33.60 | |
| $23.94 | |
| $33.60 | |
| $16.17 | |
| $63.00 | |
| $84.00 | |
| $2.63 |
Score vs price
Intelligence Index against estimated monthly token cost for your workload. Up and left is better.
- Claude Fable 5.1: 51.2 points, $210 / month (MiniRouter token estimate)
- GPT-6 Astra: 50.9 points, $210 / month (MiniRouter token estimate)
- Claude Opus 5: 48.1 points, $105 / month (MiniRouter token estimate)
- Qwen3.8 Max 0902: 45.4 points, $34 / month (MiniRouter token estimate)
- GLM 5.3: 44.8 points, $24 / month (MiniRouter token estimate)
- Grok 4.6: 44.3 points, $34 / month (MiniRouter token estimate)
- Step 5 Preview: 43.7 points, $16 / month (MiniRouter token estimate)
- Kimi K3: 43.6 points, $63 / month (MiniRouter token estimate)
Showing up to eight highest-scoring matches with a positive cost estimate. All matches are in the table.
Your selection
Claude Fable 5.1
anthropic/claude-fable-5.1Evaluated: Claude Fable 5.1 (Adaptive Reasoning, High Effort, Default Fallback) ↗
Benchmark scores describe this evaluation, not a guarantee for your prompts.
- Selected model / month
- $210.00
Same token volume at current MiniRouter rates. Actual usage, reasoning tokens, cache writes, request minimums, and routing can change the bill.
How this comparison works
Scores come from Artificial Analysis. Each task uses its own published index. Missing results stay unscored; benchmark coverage is not a measure of availability.
Only current MiniRouter catalog models appear. Open weights is a verified metadata filter, not a promise of a particular license. Confirm the model’s license before deployment.
Cost estimates use your input, output, and cache mix with MiniRouter’s catalog rates. Cache discounts apply only where a rate is published. Scores and costs come from different sources; an index point difference does not predict a percentage change in real-world quality.
Use the evaluated version and reasoning settings shown with each result when interpreting scores. Trying a model in Chat bills normal usage and may use different settings.