Loading
90 days of the enterprise, pro and team plans free, no card needed. See the plans
Model comparison
Anthropic's Claude Fable 5.1 against Moonshot AI's Kimi K3.1 Pro: vendor list prices, limits, abilities and the Artificial Analysis indexes side by side - and what OpenSmartRoute does when both are available.
Kimi K3.1 Pro lists at $4.00 per million tokens blended against $20 for Claude Fable 5.1 - 5.0x less.
Claude Fable 5.1 scores 50 on the intelligence index against 46 - 4 points apart.
Kimi K3.1 Pro reads 1.0M tokens at once, Claude Fable 5.1 1M.
Kimi K3.1 Pro was released about 2 months later.
Vendor list prices per million tokens and published limits; indexes are Artificial Analysis composites (0-100).
| Specification | Claude Fable 5.1 | Kimi K3.1 Pro |
|---|---|---|
| Input price / 1M tokens | $10.0 | $2.00 |
| Output price / 1M tokens | $50.0 | $10.0 |
| Blended (3 in : 1 out) | $20.0 | $4.00 |
| Cached input / 1M | $0.25 | $0.20 |
| Context window | 1M | 1.0M |
| Max output | 128K | 944K |
| Input | text, image, file | text, image, video |
| Output | text | text |
| Tool calls | ||
| Structured output | ||
| Reasoning | ||
| Open weights | ||
| Intelligence index | 50 | 46 |
| Coding index | 81 | - |
| Agentic index | 57 | - |
| Released | Sep 1, 2026 | Oct 23, 2026 |
| Available on this deployment |
List price times tokens for three typical workloads; your mix of input and output decides which model's pricing suits you.
| Workload | Claude Fable 5.1 | Kimi K3.1 Pro | Difference |
|---|---|---|---|
| Support chat20M tokens in, 5M out a month - a help desk answering a few thousand conversations. | $450 | $90.00 | $360 less on Kimi K3.1 Pro |
| Document summaries100M in, 5M out - long inputs, short answers. | $1,250 | $250 | $1,000 less on Kimi K3.1 Pro |
| Coding assistant60M in, 20M out - a team's editor assistant, output-heavy. | $1,600 | $320 | $1,280 less on Kimi K3.1 Pro |
Kimi K3.1 Pro takes the short and general requests - at $4.00 blended it is the default - and Claude Fable 5.1 the ones whose complexity or quality floor rules the cheaper model out. The split is per request, not per app, and every decision carries the scores for both.
# targets.yaml - both models as routing targets (metadata.model links each to the catalogue entry)
targets:
- { id: kimi-k3.1-pro, kind: llm, metadata: { model: moonshotai/kimi-k3.1-pro } }
- { id: claude-fable-5.1, kind: llm, metadata: { model: anthropic/claude-fable-5.1 } }
# rules.yaml - the cheaper model by default, the stronger one when the request is hard
rules:
- name: simple-requests-stay-cheap
when: { max_complexity: 0.45 }
prefer: [kimi-k3.1-pro]
weight: 0.8
- name: hard-requests-escalate
when: { min_complexity: 0.45 }
prefer: [claude-fable-5.1]
weight: 0.8Every Anthropic and Moonshot AI model, with prices, is on the vendor pages; the rankings show what the router actually picks.