Claude Fable 5 vs
Kimi K3

Read the published evidence, route context, and missing facts before making a local decision. This page does not name a universal winner.

Models in this comparison

Provider identity and evidence state stay balanced across the pair.

A
Model type
Proprietary
Evidence state
Supported evidence
Published context
1,000,000
VS
M

Moonshot AI

Kimi K3

Model type
Unknown
Evidence state
Supported evidence
Published context
1,050,000

Key implications

Broad shared-metric coverage. Each finding is tied to a published metric, price, context, or modality fact.

  1. Across compatible supported BenchLM categories, Claude Fable 5 has higher scores in Coding, Overall, and Agentic; Kimi K3 has a higher score in Knowledge.
  2. Input API price: Kimi K3 has the lower verified rate ($3 / 1M tokens vs $10 / 1M tokens).
  3. Output API price: Kimi K3 has the lower verified rate ($15 / 1M tokens vs $50 / 1M tokens).
  4. Context window: Kimi K3 has the larger published context window (1,050,000 tokens vs 1,000,000 tokens).

Shared metric view

The radar is a per-axis relative view; exact values and units remain available in its adjacent table.

Per-axis relative view: each axis scales to the higher published value for that exact shared metric. The table preserves the exact published values and units.
  • Claude Fable 5: solid line
  • Kimi K3: dashed line
Claude Fable 5 and Kimi K3 shared metric radarAgenticCodingKnowledgeOverall
Exact published values used in the per-axis relative radar chart.
MetricClaude Fable 5Kimi K3Unit
Agentic75.5473.64score
Coding81.4178.58score
Knowledge67.483.8score
Overall83.1580.61score

Source metrics

Friendly metric names and published units stay visible. Missing measurements remain unavailable rather than becoming a score.

Source metric comparison
MetricUnitClaude Fable 5Kimi K3
Agenticscore75.5473.64
Codingscore81.4178.58
Knowledgescore67.483.8
MultimodalGroundedscore60.886.2
Overallscore83.1580.61

Agentic

Unit
score
Claude Fable 5
75.54
Kimi K3
73.64

Coding

Unit
score
Claude Fable 5
81.41
Kimi K3
78.58

Knowledge

Unit
score
Claude Fable 5
67.4
Kimi K3
83.8

MultimodalGrounded

Unit
score
Claude Fable 5
60.8
Kimi K3
86.2

Overall

Unit
score
Claude Fable 5
83.15
Kimi K3
80.61

Pricing and context

Verification is shown beside each selected route. Missing facts remain Not verified.

Route pricing and context comparison
FieldUnitClaude Fable 5Kimi K3
Input API priceUSD / 1M tokens$10$3
Cached input API priceUSD / 1M tokens$1$0.3
Output API priceUSD / 1M tokens$50$15
Route contexttokens1,000,0001,050,000
Input modalitiespublished listNot verifiedNot verified
Output modalitiespublished listNot verifiedNot verified

Input API price

Unit
USD / 1M tokens
Claude Fable 5
$10
Kimi K3
$3

Cached input API price

Unit
USD / 1M tokens
Claude Fable 5
$1
Kimi K3
$0.3

Output API price

Unit
USD / 1M tokens
Claude Fable 5
$50
Kimi K3
$15

Route context

Unit
tokens
Claude Fable 5
1,000,000
Kimi K3
1,050,000

Input modalities

Unit
published list
Claude Fable 5
Not verified
Kimi K3
Not verified

Output modalities

Unit
published list
Claude Fable 5
Not verified
Kimi K3
Not verified

Evidence provenance

Source records, route identity, timestamps, and methodology are consolidated here without declaring either model a winner.

Publication time
Aug 30, 2026, 2:15 AM UTC
Freshness
Stale — Published weekly benchmark evidence has not refreshed within 8 days.
Methodology
benchlm: benchlm_raw_composite
Model records
  • Claude Fable 5 — source benchlm · artifact models · model claude-fable-5
  • Kimi K3 — source benchlm · artifact models · model kimi-3
Selected price routes
  • Claude Fable 5 — route benchlm:claude-fable-5 · source benchlm · provider anthropic
  • Kimi K3 — route benchlm:kimi-3 · source benchlm · provider moonshot-ai

Other reviewed matchups from the same published revision, ready to open without changing this result’s evidence.

Switch model pair

Choose from this result’s current and reviewed related models. Switching opens a reviewed comparison; it does not change this result’s evidence.

Step 1

Start with popular models, or search the full selectable directory.

Step 2

Start with popular models, or search the full selectable directory.